AU2010202702A1 - Evolving new molecular function - Google Patents

Description

Regulation 3.2 AUSTRALIA Patents Act 1990 COMPLETE SPECIFICATION DIVISIONAL APPLICATION APPLICANT: THE PRESIDENT AND FELLOWS OF HARVARD COLLEGE Invention Title: EVOLVING NEW MOLECULAR FUNCTION The following statement is a full description of this invention, including the best method of performing it known to me: WO 2004/016767 PCT/US2003/025984 EVOLVING NEW MOLECULAR FUNCTION PRIORITY INFORMATION 100011 This application claii the benefit of(i) U.S. Provisional Patent Applicaition No. 60/404,395, filed August 19, 2002, (ii) U.S. Provisional Patent Application No. 60/419,667, filed 5 October 18,2002, (iii) U.S. Provisional Patent Application No. 60/432,812, filed December 11,, 2002, (iv) U.S. Provisional Patent Application No.3,0/444,770, filed February 4, 2003, (v) U.S. Provisional Patent Application No. 60/457,789; filed March 26, 2003, (vi) U.S. Provisional Patent Application No. 60/469,866, filed May 12, 2003, and (vii) U.S. Provisional Patent Application No. 60/479,494, filed June 18, 2003, the disclosures of each of which are 10 incorporated by reference herein. The application is also related to United States Provisional Patent ApplicationNos. 60/277,081 (filed March 19, 2001), 60/277,094 (fifed March 19, 2001), 60/306,691 (filed July 20, 2001), and 60/353,565 (filed February 1, 2002), as well as to United States Patent Application Nos. 10/101,030 (filed March 19, 2002) and 10/102,056 (filed M1\arch 19, 2002), and to International Patent Application, serial number US02/08546 (filed March 19, 15 2002). GOVERNMENT FUNDING [00021 The research described in this application was sponsored, in part, by the Office for Naval Research under Contract No. N00014-00-1-0596 and Grant No. 00014-03-1-0749. The United States Government may have certain rights in the invention. 20 BACKGROUND OF THE INVENTION [0003] The classic "chemical approach" to generating molecules with new functions has been used extensively over the last century in applications ranging from drug discovery to synthetic methodology to materials science. In this approach, researchers synthesize or isolate candidate molecules, assay these candidates for desired properties, determine the structures of 25 active compounds if unknown, formulate structure-activity relationships based on available assay and structural data, and then synthesize a new generation of molecules designed to possess improved properties. While combinatorial chemistry methods (see, for example, Eliseev et al. (1999) COMBINATORIAL CHEMISTRY IN BIOLoGY 243: 159-172; Kuntz et al. (1999) CURRENT WO 2004/016767 PCT/US2003/025984 -2 OPINION IN CHEMICAL BIOLOGY 3: 313-319; Liu et al, (1999) ANGEW. CHEM. INT.ED. ENG. 38:. 36). have increased the throughput of this'approach, its fundarhental limitations remain unchanged. Several factors limit the effectiveness of the cliqmical approach to generatin molecular function. First, the ability to accurately predict the structural changes that will lead to 5 new function is often inadequate due to subtle conformational rearrangenients of molecules, unforeseen solvent interactions, or unknown stereochemical requirements of binding hr reaction eVents. The resulting complexity of structure-activity relationships frequently limits the success of rational ligand or catalyst design, including those efforts conducted in a high-throughput manner. Second, the need to assay or screen, rather than select, each member of a collection of 10 candidates limits the number of molecules that can be searched in each experiment.. Finally, the lack of a way to amplify synthetic molecules places requirements on the minimum amount of material that must be produced for characterization, screening, and structure elucidation. As a result, it can be difficult to generate libraries of more than roughly 106 different synthetic compounds. 15 [00041 In contrast, Nature generates proteins with new functions using a fundamentally different method that overcomes many of these limitations. In this approach, a protein with desired properties induces the survival and amplification of the infonnation encoding that protein. This information is diversified through spontaneous mutation and DNA recombination, and then translated into a new generation of candidate proteins using the ribosome. Unlike the 20 linear chemical approach described above, the steps used by Nature form a cycle of molecular evolution. Proteins emerging from this process have been directly selected, rather than simply screened, for desired activities. Because the biomolecules that encode evolving proteins (e.g., DNA) can be amplified, a single protein molecule with desired activity can in theory lead to the survival and propagation of the DNA encoding its structure. 25 [0005] Acknowledging the power and efficiency of Nature's approach, researchers have used molecular evolution to generate many proteins and nucleic acids with novel binding or catalytic properties (see, for example, Minshull et al. (1999) CURR. OPIN. CHEM. IOL. 3: 284 90; Schmidt-Dannert et al (1999) TRENDS BIOTECHNOL. 17: 135-6; Wilson et at. (1999) ANNU. REV. BIOCHEM. 68: 611-47). Proteins and nucleic acids evolved by researchers have 30 demonstrated value as research tools, diagnostics, industrial reagents, and therapeutics, and have greatly expanded the understanding of the molecular interactions that endow proteins and nucleic WO 2004/016767 PCTIUS2003/025984 acids with binding or catalytic properties (see, Faniulok et al (1998) CURR. OPIN. CHEM. BIOL. 2: 320-7). [00061 Despite Nature's efficient approach to generating function, Nature's ieolecular evolution is limited to two types of "natural" molecules (proteins and nucleic acids) because thus 5 far the information in nucleic acids can only he translated into proteins or into other nu6leic acids. UnfortuAately, many synthetic molecules of interest do not in general'have nucleic acid or protein backbones. An ideal approach to generating functional molecules merges the most. powerful aspects of molecular evolution with the flexibility of synthetic chemistry. Clearly, enabling the evolution of non-natural synthetic small molecules and polymers, much as Nature 10 evolves biomolecules, would lead to much more effective methods of discovering new synthetic ligands, receptors, and catalysts difficult or iinpossible to generate using rational design. 10007] ' Although these concepts have been brought together to'permit nucleit acid templated synthesis of small molecules (see, for example, Gartner & Liu (2001) 1 AM. CHEM. Soc. 123: 6961-6963) there is still an ongoing need for improvements in these core technologies 15 to permit the more efficient synthesis, selection, amplification, and evolution of molecules of interest. SUMMARY OF THE INVENTION 100081 The invention provides a variety of methods and compositions that expand the scope of template-directed synthesis, selection, amplification and evolution of molecules of 20 interest. During nucleic acid-templated synthesis, the information encoded within a nucleic acid template is used to bring two or more reactants together into reactive proximity. These methods permit the creation of, for example, small molecule and polymer libraries that have not been possible to create to date using conventional combinational chemistries. [00091 In one aspect, the invention provides a method of performing nucleic acid 25 templated synthesis using a template having an "omega" or " 9L" type architecture. This type of template permits distance-dependent nucleic acid-templated reactions to be encoded by bases far removed from the associated reactive unit. The method involves providing (i) a template comprising a first reactive unit associated with a first oligonucleotide comprising a codon and (ii) a transfer unit comprising a second reactive unit associated with a second oligonucleotide 30 comprising an anti-codon that is capable of annealing to the codon. The codon and/or the anti- WO 2004/016767 PCTIUS2003/025984 -4 codon include first and second regions spaced apart from one another. The oligonucleotides then are annealed together to bring the reactiVe units into reactive proximity. When the oligonucleotides anneal to one another; the codon (or anti -o&don) with the spaced-apart regions produce a loop of oligonucleotides not annealed to the corresponding anti-codon (or codon). A 5 covalent bond-forming reaction then is induced between the reactive units to produce the reaction product. [00101 In one embodiment, at least one of thd reactive units are attached adjacent a terminal region of its corresponding oligonucleotide. In another embodiment, the codon or anti codon is disposed more than one base away (for example, 10, 20, 30 bases or more) from its 10 corresponding reactive unit. The first spaced apart region typically is disposed directly adjacent a terminus of its corresponding oligonucleotide. The first spaced apart region preferably includes, for example, three, fonr, or five nucleotides, although other embodiments (e.g., more than five nucleotides) are also envisione. The second region may be disposed, for example, at least twenty or at least thirty bases away from its corresponding reactive unit. More particularly, 15 the end of the second region closest to the reactive unit may be disposed, for example, at least ten, twenty, thirty or more bases from the end of the oligonucleotide attached to its reactive unit. The template may include additional (e.g., 2, 3, 4, or more than 4) codons, in which case a corresponding number of transfer units can be annealed to the template, optionally permitting multi-step or alternative syntheses. 20 [0011] In another aspect, the invention provides a method of performing a nucleic acid templated synthesis using a template having a "T" type architecture. The T architecture permits two nucleic acid-templated reactions to take place on a single template in a single step. The method involves providing (i) a template comprising a first reactive unit (e.g., a scaffold molecule) associated with a first oligonucleotide having a codon; and (ii) a transfer unit 25 comprising a second reactive unit associated with a second oligonucleotide having an anti-codon capable of annealing to the codon. The first reactive unit is attached, preferably covalently, to an attachment site intermediate the proximal and distal ends of the first oligonucleotide of the template. During synthesis, the oligonucleotides of the template and transfer unit are annealed to one another to bring the reactive units into reactive proximity, and a covalent bond-forming 30 reaction between the reactive units is induced.

WO 2004/016767 PCT/US2003/025984 -5 [00121 In one embodiment of the T type architecture,.the template also includes a second. different codon capable of annealing to second, different anti-codon sequence of a second, different transfer unit. In this embodirent, the first codon is located proximal to the attachment site and the second codon, if present, is located distal to the attachment site. If a second transfer 5 unit comprising a third reactive unit associated with a third oligonucleotide having a second, different anti-codon sequence capable of annealing to. the second codon is provided, the second transfer unit may bind to the template at the second codon position. Accordingly when the first 'and second transfer units are combined with the template, the first anti-codon of the first transfer unit anneals to the first codon of the template and the second anti-codon of the second transfer 10 unit anneals to the second codon of the template. This system permits two reactions to occur simultaneously or sequentially on a single template in a single step. [00131 In another aspect, the invention provides a series of methods for increasing reaction selectivity between reactants in a templated synthesis. In one approach, the method comprises providing a template and at least two transfer units. The template comprises a first 15 reactive unit associated with a first oligonucleotide comprising a predetermined codon sequence. The first transfer unit comprises a second reactive unit associated with a second oligonucleotide comprising an anti-codon sequence capable of annealing to the codon sequence. The second transfer unit comprises a third reactive unit, different from the second reactive unit. The third reactive unit, however, is associated with a third oligonucleotide that lacks an anti-codon 20 sequence capable of annealing to the codon sequence. The template and transfer units are mixed under conditions to permit annealing of the second oligonucleotide to the first oligonucleotide, thereby to enhance covalent bond formation between the second and first reactive units relative to covalent bond formation between the third and first reactive units. [00141 This method may be particularly helpful when the second and third reactive units 25 are each capable of reacting independently with the first reactive unit. Furthermore, the method may also be helpful when the second and third reactive units are capable of reacting with one another, for example, to modify or inactivate one another. Accordingly, this type of method permits a series of otherwise incompatible reactions to occur in the same solution, for example, where a reaction between the second and third reactive units is incompatible with a reaction 30 between the second reactive unit and the first reactive unit. The method may enhance covalent bond formation between the first and second reactive units by at least 2-fold, at least 5-fold, at WO 2094/016767 PCTf/JS2003/025984 -6 least I 0-fold, or at least 50-fold relative to covaentbond formation between the first and third reactive units. Collectively, these advaritages permit a ohe-pot ordered multi-step synthdis in which a sequence of reactions is programmed by the seqpence of a templai6 oligonueleotide. Thus, a sequence of atleast 2, 3, 4, 5, 6, or more reactions can take place in an ordered manner in 5 a single solution, eveh when the reactants would interfere with each other using conventional, non-templated dheristries. . I [00151 In one embodiment, the template, the first transfer unit, and/or the second transfer. unit are associated with a capturable moiety; for example, biotin, avidin, or streptavidin. If a. capturable moiety is present, the method may include capturing the capturable moiety as a way 10 to enrich a reaction product from a reaction mixture. 10016] In another approach, the method comprises providing (i) a template comprising a first oligonnlcleotide having first and second coden sequences (ii) a first transfer unit, (iii) a second transfer unit, and (iv) a third transfer unit. The first transfer unit comprised a first reactive unit associated with a second oligonucleotide comprising a first anti-codon sequence 15 capable of annealing to the first codon sequence. The second transfer unit comprises a second reactive unit associated with a third oligonucleotide comprising a second anti-codon sequence capable of annealing to the second codon sequence. The third transfer unit comprises a third reactive unit associated with a fourth oligonucleotide sequence that lacks an anti-codon sequence capable of annealing to the first or second codon sequences. The template, first transfer unit, 20 second transfer unit, and third transfer unit then are mixed under conditions to permit (i) annealing of the first anti-codon sequence to the first codon sequence and (ii) annealing of the second anti-codon sequence to the second codon sequence thereby to enhance covalent bond formation between the first and second reactive units relative to covalent bond formation between the third reactive unit and the first reactive unit and/or between the third reactive unit 25 the second reactive unit. This type of method may be particularly useful for producing non natural polymers by nucleic acid-templated synthesis. 100171 In one embodiment, the template is associated with a capturable moiety, for example, biotin, avidin, or streptavidin. The capturable moiety may also be a reaction product resulting from a reaction between the first and second reactive units when the first and second 30 reactive units are annealed to a template. If a capturable moiety is present, the method may WO 2004/016767 PCT/US2003/025984 -7 include capturing the capturable moiety as a way to enrich a reaction production from the reaction mixture. [00181 This type of method is also helpful when the third reactive unit is capable of reacting with the first and/or second reactive units. In other words, the reaction between the first 5 and third reactive units and/or between the second and third reactive units may be incompatible with the reaction between the first and secon' reactive units. The method may enhance covalent bond formation between the first and second reactive units by at least 2-fold, at last 5-told, at least 10-fqld, or at least 50-fold relative to covalent bond formation between the first and third reactive units. 10 [0019] In another aspect, the invention provides a series of methods for performing stereoselective nucleic acid-templated synthesis. The stereoselectivity of the synthesis may result fromthe choice of a particular template, transfer unit, reactive unit, hybridized template and transfer unit, stereoselective catalyst, or any combination of the above. The resulting product may be at least 60%, at least 70%, at least 80%, at least 90%, at least 95%, at least 98%, 15 or at least 99% stereochemically pure. [00201 Generally, the method involves providing (i) a template comprising a first oligonucleotide that optionally is associated with a reactive unit and (ii) one or more transfer units, each comprising a second oligonucleotide associated with a reactive unit. Annealing-of the first and second oligonucleotides brings at least two reactive units into reactive proximity and 20 to react to produce a reaction product where the reaction product contains a chiral center and is of at least 60%, more preferably at least 80%, and more preferably at least 95% stereochemically pure at the chiral center. It is contemplated that this method can be accomplished when one reactive unit is associated with the template and the other reactive unit is associated with the transfer unit. Also, it is contemplated that this method can be accomplished when the template 25 does not provide a reactive unit and two transfer units when they anneal to the template provide the two reactive units that come into reactive proximity to produce the reaction product. [00211 In one approach, the method involves providing at least two templates and at least one transfer unit. One template includes a first oligonucleotide associated with a first reactive unit comprising a first stereochemical configuration, and the other template includes another first 30 oligonucleotide associated with another first reactive unit having a second, different stereochemical configuration. The transfer unit comprises a second reactive unit associated with WO 2004/016767 PCT/US2003/025984 -8 a second oligonucleotide including a sequence complementary to a sequence of the first oligonucledtide of the template. The first and, second oligonucleotides then are annealed under conditions to permit the second reactive unit of the transfer unit'to react p'eferentially with either the first reactive unit of the first stereochemical configuration or the first reactive unit of the 5 second stereochemical - configuration to produce a reaction product. [0022] The resulting reaction product may have-a particular stereochemical configuration. In one embodiment, a stereochemical configuration or macromolecular conformation of the first oligonucleotide of the template determines whidh one of the first reactive units reacts with the second reactive unit. 10 [0023] In a second approach, the method involves providing at least one template and at least two transfer units. The template includes a first oligonucleotide associated with a first reactive unit. One transfer unit comprises' a second oligonucleotide associated with a second reactive unit having a first stereochemical configuration, and the other transfer unit comprises another second oligonucleotide associated with a second reactive unit having a second, different 15 stereochemical configuration. A sequence of the second oligonucleotides is complementary to a sequence of the first oligonucleotide. The first and second oligonucleotides then aie annealed under conditions to permit the first reactive unit of the template to react preferentially with either the second reactive unit having the first stereochemical configuration or with the second reactive unit having the second stereochemical configuration to produce a reaction product. 20 100241 The resulting reaction product may have a particular stereochemical configuration. In one embodiment, a stereochemical configuration or macromolecular conformation of the second oligonucleotide determines which of the second reactive units reacts with the first reactive unit. [0025] In a third approach, the method involves providing at least one template and at 25 least two transfer units, wherein one or optionally both of the transfer units comprise a pair of reactive units with one reactive unit of the pair having a first stereochemical configuration and the other reactive unit of the pair having a second, different stereochemical configuration. The template comprises a first oligonucleotide comprising a first codon sequence and a second codon sequence. One transfer unit of a first pair of transfer units includes a second oligonucleotide 30 with a first anti-codon sequence associated with a first reactive unit having a first stereochemical configuration. The other transfer unit of the first pair of transfer units includes another second WO 2004/016767 PCT/US2003/025984 -9 oligonucleotide associated with a second stereochemical configuration of the first reactive unit. The second transfer unit includes a third oligonucleotide with'a second anti-codon sequence associated with a second reactive unit. The template, the fist pair of transfer units; and the second transfer unit are annealed to permit a member of the first pair of transfer units to react 5 preferentially with the second transfer unit to produce a reaction product. The resulting reaction product may have a particular stereochemical configuration. 10026] In one embodiment, a stereochemical'donfiguration or macromole'cular conformation of the second oligonucleotide determines which member of the first.pair of transfer units reacts preferentially to produce the reaction product. 10 [00271 In one embodiment, the method involves providing a template and at least two pairs of transfer units. The template comprises a first oligonucleotide comprising first and second codon sequences. One transfer unit of the first pair comprises a second oligonuclcotide with a first anti-codon sequence associated with a first reactive unit having a first stereochemical configuration. The other transfer unit of the first pair comprises the second oligonuckotide with 15 the first anti-codon sequence associated with a first reactive unit having a second, different stereochemical configuration. One transfer unit of the second pair of transfer units comprises a third oligonucleotide having a second, different anti-codon sequence associated with a second reactive unit having a first stereochemical configuration. The other transfer unit of the second pair comprises the third oligonucleotide with the second anti-codon sequence associated with the 20 second reactive unit having a second, different stereochemical configuration. The template, the first pair of transfer units and the second pair of transfer units are annealed to permit a member of the first pair of transfer units to react preferentially with a member of the second pair of transfer units to produce a reaction product. 100281 In one embodiment, a stereochemical configuration or macromolecular 25 conformation of the second oligonucleotide determines which member of the first pair of transfer units reacts preferentially to produce the reaction product. In addition, a stereochemical configuration or macromolecular conformation of the third oligonucleotide determines which member of the second pair of transfer units reacts preferentially to produce the reaction product. [0029] In another aspect, the invention provides a method for enriching a product of a 30 templated synthesis reaction. The method comprises providing a first library of molecules comprising a plurality of reaction products associated with a corresponding plurality of WO 2004/016767 PCT/US2003/025984 oligonucleotides, wherein each oligonucleotide comprises a nucleotide sequence indicative of the, associated reaction product. A portion df the reaction products in the first library are capable of binding to a preselected moiety. The'first library their is exposed to the binding moiety under conditions to permit reaction product capable of binding the binding moiety to do so. Unbound 5 reaction products are removed, and bound reaction product then is eluted froni the binding moiety to produce a second library of molecules enriched at least 10-fold, mere preferably at' least 50-fold, relative to the first library, for ration products that bind the binding moiety. 100301 In one eftbodiment, the binding moiety, for example, a targqt biomolecul& for example, a protein, is immobilized on a solid support. hn another embodiment, the second 10 library is enriched at leas1' 100-fold or at least I '000-fold for reaction products that bind to the binding moiety. Furthermore, it is contemplated that the steps of exposing the library to the binding mojety, removing unbound reaction products, and eluting bound reaction prpducts can be repeated (eg., repeated one, two, three or more times) Repetition of these steps preferably yields a second library enriched at least 1,000-fold, more preferably, at least 10,000-fold, or, 15 more preferably, at least 100,000-fold, for reaction products that bind to the binding moiety. 100311 In one embodiment, the oligonucleotide attached to the selected library member includes a first sequence that identifies a first reactive unit that produced the reaction product bindable by the preselected binding moiety. Preferably, the oligonucleotide also includes a second sequence that identifies a second reactive unit that produced the reaction product 20 bindable by the preselected binding moiety. By sequencing the oligonucleotide attached to the selected library member it is possible to determine what reactants reacted with one another to produce the reaction product. Accordingly, using this approach it is possible to deduce the structure of the selected library member from the reaction history. [00321 The method may further comprise the step of amplifying the oligonucleotide 25 associated with the enriched reaction product and, preferably, determining the sequence of the amplified oligonucleotide. Furthermore, the reaction product can be further characterized by using information encoded within the sequence of the oligonucleotide. For example, the sequence of the oligonucleotide may be determined and then from the sequence it is possible to determine what reactive units reacted to produce the reaction product. Using a similar approach, 30 it is possible to identify the existence of new chemical reactions that produced the reaction product.

WO 2004/016767 PCT/US2003/025984 [00331 in another aspect, the invention provides a variety of methods for identifying the existence of new chemical reactions. One approach involves, providing a library of moldcules ' comprising a pluraifty of reaction products associated with a corresponding plurality of oligonbcleotides, wherein each oligonucleotide includes 'a nucleotide sequence indicative of an 5 associated'reaction product. A particular reaction product associated with its 6orrespozding oligonucleotideihen is selected, and characterized, Following characterization of the reaction product and identification of the reactive units hat reacted to create the reaction product, it ig possible to identify one or more new Chemical reactiops necessary to produce the reaction product. 0 100341 In one embodiment, the method further includes, after selecting the reaction product, amplifying its corresponding oligoinucleotide. The amplified oligonucleotide can then be sequenced to identify what reactive units reacted to produce the reaction product, The oligonucleotid& may also be amplified for use in preparing more of the selected reaction product In other embodiments, the oligonucleotide niay be mutated, and the resulting mutated 5 oligonucleotide may be used in the creation of a second generation library. [0035] A second approach involves providing (i) a template and (ii) a first transfer unit. The template comprises a first reactive unit associated with a first oligonucleotide comprising a codon. The transfer unit comprises a second reactive unit associated with a second oligonucleotide comprising an anti-codon capable of annealing to the codon. The 0 oligonucleotides are annealed to bring the first and second reactive units into reactive proximity. A covalent bond-forming reaction is induced between the reactive units to produce a reaction product. The reaction product then is characterized, and a new chemical reaction necessary to make the reaction product is identified using information encoded by the template to identify the first and second reactive units that reacted to produce the reaction product. The method may also 5 include the step of selecting the reaction product prior to its characterization. [0036] In a third approach, the invention involves providing at least (i) a template, (ii) a first transfer unit and (iii) a second transfer unit. The first transfer unit comprises a first reactive unit associated with a first oligonucleotide. The second transfer unit comprises a second reactive unit associated with a second oligonucleotide. The template includes sequences capable of 30 annealing to the first and second oligonucleotides. During the method, the oligonucleotides are annealed to the template to bring the reactive units into reactive proximity and a covalent bond- WO 2004/016767 PCT/US2003/025984 -12 forming reaction is induced between the reactive units to produce a reaction product The reaction product then is characterized, for example, by using information encoded by the template to identify the first and second reactive units thai reacted with one another to produce the reaction product. Based on the characterization, it is then possible to identify one or more 5 new chemical reactions that were necessary to make the reaction product. The method may also include the step of selecting the reaction product prior to its characterization. 100371 Although the methods of the invention are useful with small numbers of templates and transfer units, use of larger numbers of templates (ag., 10, 50, 100, 1000, ormore) and of transfer units for each codon (e.g., 10, 20, 30, 50, or more) permits the synthesis of large libraries 10 of molecules that can be screened simultaneously using the sensitivity afforded by amplification. DEFINITIONS [00381 The term, "associated with" as used herein describes the interaction between or among two or more groups, moieties, compounds, monomers, etc. When two or more entities are "associated with" one another as described herein, they are linked by a direct or indirect 15 covalent or non-covalent interaction. Preferably, the association is covalent. The covalent association may be, for example, but without limitation, through an aide, ester, carbon-carbon, disulfide, carbamate, ether, thioether, urea, amine, or carbonate linkage. The covalent association may also include a linker moiety, for example, a photocleavable linker. Desirable non-covalent interactions include hydrogen bonding, van der Waals interactions, dipole-dipole 20 interactions, pi stacking interactions, hydrophobic interactions, magnetic interactions, electrostatic interactions, etc. Also, two or more entities or agents may be "associated with" one another by being present together in the same composition. [00391 The term, "biological macromolecule" as used herein refers to a polynucleotide (e.g., RNA, DNA, RNA/DNA hybrid), protein, peptide, lipid, or polysaccharide. The biological 25 macromolecule may be naturally occurring or non-naturally occurring. In a preferred embodiment, a biological macromolecule has a molecular weight greater than about 5,000 Daltons. [0040] The terms, "polynucleotide," "nucleic acid", or "oligonucleotide" as used herein refer to a polymer of nucleotides. The polymer may include, without limitation, natural 30 nucleosides (i.e., adenosine, thymidine, guanosine, cytidine, uridine, deoxyadenosine, WO 2004/016767 PCT/US2003/025984 - 13 deoxythyrnidine, deoxyguanosine; and deoxycytidine),7ucleoside analogs (eag, 2 aminoadenosine, 2-thiothymidine, inosine, yyrrojQ-pyrimidine, 3-methyl adenosine, 5 methylcytidine, 'C-bromouridine, C 'fluorduridine, C5-iodouridihe, C5-yiopynyl-uridine, CS-propynyl-cytidine, C5-methyleytidine, 7-deizaadenosine, 7-deazaguanosine, 5 8-oxoadenosine, 8-oxoguanosine, O(6)-methylguanine, and 2-thiocytidine), chemicallymodified bases, biologically modified bases (e.g., methylated. bases), intercalated bases, modified, sugars (e.g., 2'-fluororibose, ribose, 2'-deoxyribose,,.arabinos, and hexose), or modified phosphate groups (tg., phosphorothioates and-5' -N-phosphoramidite linkages). Nucleic acids and. oligonucleotides may also include other polymers of bases having a modified backbone, such as 10 a locked nucleic acid (LNA), a peptide nucleic acid (P-NA), a threose nucleic acid (TNA) and any other polymers capable of serving as a template for'an amplification reaction using an amplification technique, for example, a polyinerase chain reaction, a ligase chain reaction, or non-enzymatictemplate-directed replication. [0041] . The tenn, "small molecule" as used herein,.refers to an organic compound either 15 synthesized in the laboratory or found in nature having a molecular weight less than 10,000 grams per mole, optionally less than 5,000 grams per mole, and optionally less than 2,000 grams per mole. [0042] The terms, "small molecule scaffold" or "molecular scaffold" as used herein, refer to a chemical compound having at least one site or chemical moiety suitable for 20 functionalization. The small molecule scaffold or molecular scaffold may have two, three, four, five or more sites or chemical moieties suitable for functionalization. These functionalization sites may be protected or masked as would be appreciated by one of skill in this art. The sites may also be found on an underlying ring structure or backbone. [0043] The term, "transfer unit" as used herein, refers to a molecule comprising an 25 oligonucleotide having an anti-codon sequence associated with a reactive unit including, for example, but not limited to, a building block, monomer, monomer unit, molecular scaffold, or other reactant useful in template mediated chemical synthesis. [0044] The term, "template" as used herein, refers to a molecule comprising an oligonucleotide having at least one codon sequence suitable for a template mediated chemical 30 synthesis. The template optionally may comprise (i) a plurality of codon sequences, (ii) an amplification means, for example, a PCR primer binding site or a sequence complementary WO 2004/016767 PCT/US2003/025984 -14 thereto, (iii) a reactive unit associated therewith, (iv) a combination of (i) and (ii), (y)- a combination of (i) and (iii), (vi) a combination of (ii) and (iii), or a combination of (i), (ii) and (iii). [0045] The terms, "codon" and "anti-codon" as used herein, refer to complementary 5, oligonucleotide sequences in the template and in the transfer unit, respectively, that permit the transfer unit to anneal to the template during iemplat mediated chemical synthesis. [0046] Throughout the description, where compositions are described as having, including,'or comprising specific components, or where processes are described as'having, including, or comprising specific process steps, it is contemplated that compositions of the 10 present invention also consist essentially of, or consist of, the recited components, and that the processes of the present invention also consist essentially of, or consist of, the recited processing steps. Further, it should be understood that the order of steps or order for performing certain actions are immaterial so long as the invention remains operable. Moreover, two or more steps or actions may be conducted simultaneously. 15 DESCRIPTION OF THE DRAWINGS [0047] Figure 1 depicts known sequence-specific oligomerizations of complimentary oligonucleotides catalyzed by single-stranded nucleic acid templates. [0048] Figure 2 is a schematic representation of one embodiment of nucleic acid templated synthesis where a reactive unit is attached to a template at the start of synthesis. 20 [0049] Figure 3 is a schematic representation of a second embodiment of nucleic acid templated synthesis where a reactive unit is not attached to the template at the start of synthesis. [0050] Figure 4 is a schematic representation of a third embodiment of nucleic acid templated synthesis suitable for polymer synthesis. [0051] Figures 5A-F are schematic representations of various exemplary templates 25 useful in nucleic acid-templated synthesis. [0052] Figures 6A-E are schematic representations of desirable and undesirable possible interactions between a codon of a template and an anti-codon of a transfer unit. [0053] Figures 7A-G are schematic representations of various template architectures useful in nucleic acid-templated synthesis.

WO 2004/016767 PCT/US2003/025984 - 15 [00541 Figure 8 is a schematic representation of a method for producing a template, containing, from the 5'-end to the 3'-end, a-small moledule functional group; a DNA hairpin, an annealing region, a oding region, and a PCR primer binding site. [00551. Figure 9 is a schematic representation of a general method for making a library of 5 reaction products. 10056] Figure, 1 is'a graph showing thq relationship between the effective concentration of target protein and the fraction of ligand thatbinds the target. {0057] Figures 11 A-B are schematic represntations of methods for screening a library for bond-cleavage (Figure 11A) and bond-formation (Fighre IB) catalysts. 10 [0058] Figure 12'is a schematic representation of an in vitro selection scheme for identifying non-natural polymer catalysts of bdnd-orming reactions. [0059] Figure 13 is a schematic representation of an in vitro selection scherne for identifying non-natural polymer catalysts of bond-cleaving reactions. [0060] Figure 14 is a schematic representation of exemplary reagents and their us6 in a 15 recombination method for diversifying a template library. [00611 Figure 15 depicts synthetic reactions directed by hairpin (H) and end-of-helix (E) DNA templates. Reactions were analyzed by denaturing polyacrylamide gel electrophoresis (PAGE) after the indicated reaction times. Lanes 3 and 4 contained templates quenched with excess p-n'ercaptoethanol prior to reaction. 20 [0062] Figure 16 depicts the results of reactions between matched (M) or mismatched (X) reagents linked to thiols (S) or primary amines (N) and templates functionalized with the variety of electrophiles. [0063] Figures 17A-17B depict various mismatch reactions analyzed by denaturing PAGE. Figure 17A depicts results of reactions in which H templates linked to an iodoacetamide 25 group were reacted with thiol reagents containing 0, 1, or 3 mismatches at 25 0 C. Figure 17B depicts results of reactions in which the reactions in Figure 17A were repeated at the indicated temperatures for 16 hours. {00641 Figure 18 depicts a reaction performed using a 41-base E template and a 10-base reagent designed to anneal 1-30 bases from the 5' end of the template.

WO 2004/016767 PCT/US2003/025984 -16 [00651 Figure 19 depicts a repeat of then =.10 reaction in Figure 18 in which the nine bases following the 5'-NH2-dT were replaced with various backbone analogues. [0066] Figure 20 depicts the n = 1, n = 10, and n = 1 mismatched (mis) reactions described in Figure 18 which were repeated with templatb'and reagent concentrations of 12.5, 5 25, 62.5 or 125 nM. [00671 Figures 21A-211B are a schematic reorbsentation of a method for translating, selecting, and amplifying a synthetic molecule that binds streptavidin from a DNA-encoded' library. [00681 Figure 22A depicts DNA sequencing results of a PCR amplified poQl of nucleic 10 acid templates of Figures 21A-21B before and after selection. [00691 Figure 22B is a schematic representation of a method for creating and evolving libraries of non-natural molecules using hucleic acid-templated synthesis, where -R, represents the library of product functionality transferred from reagent library 1 and -R 8 represents a selected product. 15 [00701 Figures 23A-23D are schematic representations of exemplary DNA-templated reactions. [0071] Figure 24 depicts analysis by denaturing PAGE of representative DNA-templated reactions listed in Figures 23 and 25. [00721 Figures 25A-25B are schematic representations of DNA-templated amide bond 20 formation reactions mediated by EDC and sulfo-NHS or by DMT-MM for a variety of substituted carboxylic acids and amines. 100731 Figure 26A-26B depict an analysis of the distance independent nature of certain nucleic acid-templated reactions. Figure 26A is a schematic representation showing a model for distance-independent nucleic acid-templated synthesis. Figure 26B depicts the results of 25 denaturing PAGE of a DNA-templated Wittig olefination between complementary aldehyde linked template 11 and phosphorous ylide reagent 13 from Figure 23B with either zero bases (lanes 1-3) or ten bases (lanes 4-6) separating annealed reactants. [00741 Figure 27 is a schematic representation of exemplary nucleic acid-templated complexity building reactions.

WO 2004/016767 PCT/US2003/025984 -17 [00751 Figures 28A-28B depict strategies for DNA-!templated synthesis using autocleaving linkers (Figures 28A and 28B), scarless linkers (Figure 28C), and useful scar linkers (Figure 28D). - [0076] Figure 29 depicts results from nucleic acit--templated reactions with various 5 linkers. [0077] Figures 30A-30B are schematic representations depicting strategies for purifying products of DNA-templated synthesis using an autocleaving reagent linker (Figure 30A) or scar and non scar linkers (Figure 30B). [0078] Figures 3 1A-B depict an'exemplary DNA-templated multi-step tripeptide 10 synthesis. [0079] Figures 32A-B depict an exemplarf4DNA-templated multi-step synthesis. [0080] Figure 33 depicts DNA-{emplated aide bond formation reactions in which reagents and templates are complexed with dimethyldidodecylammonium cations. [0081] Figure 34 shows denaturing PAGE gels with representative DNA-templated 15 amine acylation, Wittig olefination, 1,3-dipolar cycloaddition, and reductive amination reactions using the end-of-helix (E) and omega (Q) architectures. 100821 Figures 35A-35D are bar charts showing a comparison of end-of-helix (E)' hairpin (H), and omega (0) architectures for mediating DNA-templated amine acylation (Figure 35A), Wittig olefination (Figure 35B), 1,3-dipolar cycloaddition (Figure 35C), or reductive 20 amination reactions (Figure 35D). [0083J Figure 36 is a table showing the melting temperatures of selected template reagent combinations using the omega (0) and end-of-helix (E) architectures. [0084] Figure 37 is a bar chart showing the efficiencies of DNA-templated reactions mediated by a template having the T architecture. 25 [0085] Figures 38A-38C depict two DNA-templated reactions on a single template in one solution mediated by templates having a T architecture. [0086] Figure 39A-39C are schematic illustrations showing the relative rates of product formation from (S)-and (R)-bromides in H template (Figure 39A) or E template (Figures 39B and 39C) mediated stereoselective DNA-templated substitution reactions.

WO 20041016767 PCT/US20031025984 - 18.

[00871 Figures 40A-40D depict results on reaction stereoselectivity when aromatic bases between the reactive groups are deleted'and restored; The Figures show changes in stereoselectivity as 'a result of restoring aromatic DINA bases fr6m the 5' end (Figures 40A-40C) or from the 3' end (Figure 40D) of the 12-basd intervening region. 5 [00881 Figu'res 41A-41B show the streohelectivities of DNA-templated reactions mediated by riht-handed helix (B-form) (Figure 41A) of left-handed helix (Z-forn) (Figures . 41A and 41,B) hairpin architectures. [00891 Figures 42A-42D shows graphical prepresentations of product yield versus time for exemplary stereoselective DNA-templated'reactions used to calculate ksR. Figure 42A' 10 corresponds to the reaction shown in Figure 39A; Figure 42B corresponds to the reaction shown in Figure 39B; Figure 42C corresponds to thereaction shown in Figure 44A and Figure 42D corresponds to the reaction shown in Figure,44B. 100901 Figures 43A-43F are a schematic representations showing template' and reagent structures that incorporate achiral, flexible linkers. 15 [00911 Figure 44A-44B are graphical representations of circular dichroism spectra obtained for B-form (Figure 44A) and Z-form (Figure 44B) template-reagent complexes. [0092] Figure 45 shows a representative denaturing PAGE analysis of reactions using the CG-rich sequences at low and high salt concentrations. [0093] Figure 46 is a schematic representation of a DNA-templated synthesis in which 20 maleimides, aldehydes, or amines are subjected to multiple DNA-templated reaction types in a single solution. [0094] Figure 47 depicts templates and reagents used pairwise in 12-reactant one-pot DNA-templated reactions. [0095] Figure 48 depicts a "one-pot" DNA-templated reaction containing 12 reactants 25 and at least seven possible reaction types which generates only 6 sequence-programmed products out of at least 28 possible products. [0096J Figure 49 is a schematic representation of a method for diversifying a DNA templated library by sequentially exposing or creating reactive groups.

WO 2004/016767 PCT/US2003/025984 -19 [00971 Figures 50A-50E are schematic representations of exemplary inwoisaid templated deprotections useful in the practice of the invention. [00981, Figures 51A-51B are schematic representations of exemplary nucleic acid templated functional group interconversions useful in the practice of the invention. .5 [00991 Figure 52 is a schematic, representation showing the assembly of transfer units along a nucleic acid template. 101001 Figure 53 is a schematic representation showing the polymerization of dicarbanate units along a nucleic acid template to form a polycarbamate. [01011 Figure 54 is a schematic representation showing cleavage of a polycarbamate 10 polymer from a nucleotide backbone. [01021 Figure 55 is a schematic representation showing the synthesis of a DNA templated macrocyclic fumaramide library. [0103] Figure 56 is a schematic representation of the amine acylation and cyclization steps of various fumaramide library members of Figure 55. 15 [0104] Figure 57 shows exemplary amino acid building blocks for the synthesis of a DNA-templated macrocyclic fumaramide library. [0105) Figure 58 is a schematic representation of a method of creating a template used in the synthesis of a DNA-templated macrocyclic fumaramide library. [0106] Figure 59 is a schematic representation of an amine acylation and cyclization 20 reaction useful in the synthesis of macrocyclic fumaramide library. [0107] Figure 60 depicts representative monomer structures that can be incorporated into a PNA polymer. [0108] Figure 61 is a schematic representation of a method for making functional polymers. As shown the polymer is still associated with the template. 25 [0109] Figure 62 depicts a DNA-templated aldehyde polymerization reaction. [0110] Figure 63 depicts PNA polymerization reactions using a 40 base template with mismatched coons located at certain positions of the template. [0111] Figure 64 shows the specificity of DNA-templated polymerization reactions.

WO 2004/016767 PCT/US2003/025984 -20- 101121 Figure 65A is a schematic representation showing a method of using a nucleic acid to direct the synthesis of new polymers apd plastics. Figure 65B is a schematic representation sho)ding the use of Gribbs' fing-openingmetathesis polyrifefization catalysis to evolve plastics. 5 [01131 Figu're 66 is a schematic represenfationshowing the evolution of plastics through iterative cycles of ligand diversification, selection, and aniplification to create polymers witl desired properties. [0114] Figure 67 depicts exemplary functionalized nucleotides that can be incorporated by DNA polymerase. 10 [01151 Figure 68 depicts exemplary metal binding uridine and 7-deazaadenosine analogs. [0116] Figure 69 depicts an exemplary synthesis of analog 7 from Figure 67. [01171 Figure 70 depicts an exempiary' synthesis of compound 30, a precursor to, compound 13 from Figure 67. 15 101181 Figure 71 depicts an exemplary synthesis of compound 40, a precursor to compound 13 from Figure 67. 101191 Figure 72 depicts an exemplary synthesis of compound 38, a precursor to compound 40 from Figure 71. [01201 Figure 73 depicts exemplary deoxyadenosine derivatives. 20 [0121] Figure 74 depicts an exemplary synthesis of modified deoxyadenosine triphosphates. [01221 Figure 75 depicts a summary of modified nucleotide triphosphates containing metal-binding functionalities which are or are not incorporated by DNA-polymerase. [0123] Figure 76 depicts a non-natural polymer library containing a synthetic metal 25 binding nucleotide that is compatible with DNA polymerases. [0124] Figure 77 is a schematic representation showing the generation of libraries of nucleic acids containing polymerase-accepted metal-binding nucleotides.

WO 20041016767 PCT/US2003/025984 -21 [01,25] Figures 78A-78C show reaction schemes for identifying certain reaction ctalysts. Figure 78A is a schematic representation of an exemplary scheme for the in vitro selection of synthetic polymers containiAg polynerase-ac'epted metal-binding nucleotides that catalyze Heck reactions. Figure 78B is a schematic representation of an exemplary scheme for 5 the in vitro selection of synthetic polymers containing polymerase-accepted metal-binding nucleotides that catalyze hetero Diels-Afder reactions. Figure 78C is a schematic representation of an exemplary scheme for the in vitro selection of synthetic polymers containing polymerase accepted metal-binding nucleotides that catalyze aldol reactions. [0126] Figure 79 depicts exemplary DNA-linked synthetic molecules subjected to 10 protein binding selections, and enrichment factors for a single round of selection. [0127] Figure 80 depicts the results of an exemplary selection scheme. [0128] Figure 81 depicts the net enrichment realized by three rounds of enrichment. [0129] Figure 82 depicts the separation of target-specific and non-specific DNA-linked synthetic molecules from a single solution. 15 [0130] Figure 83 depicts exemplary specific DNA-linked synthetic molecules selected in Figure 79. [01311 Figure 84 depicts an exemplary iterated carbonic anhydrase selection scheme. .[01321 Figure 85 is a schematic representation of a method for performing one-pot selections for bond-forming reactions. 20 [0133] Figure 86 is a schematic representation of a method for validating the discovery of new bond-forming reactions using DNA-templated synthesis. [0134] Figure 87 depicts an example of reaction discovery using nucleic acid-templated synthesis. [0135] Figure 88 depicts the discovery of Cu-mediated coupling reactions identified 25 using nucleic acid-templated synthesis. [0136] Figure 89 depicts the discovery of Pd-mediated coupling reactions identified using nucleic acid-templated synthesis. [0137] Figure 90 is a schematic representation of a microarray based sequence analysis protocol.

WO 2004/016767 PCTIUS2003/025984 1 -22 [0138] Figure 91 depicts the analysis "of tht Pdpmediated reactions identified via microarray based sequence analysis ' DESCRIPTION OF CERTAIN EMBODIMENTS OF THE INVENTION [0139]' 'Nucleic-acid templated synthesis as described herein permits the production, 5 selection, amplification and evolution, of a brad Varipty' of chemical compounds such as synthetic small molecules and non-natural polymers.. In nucleic acid-templated synthesis; the information 'encoded by a DNA or other nucleid acid sequence is translated into the synthesis of a reaction product. The nucleic acid template typicall'comprises a plurality of coding regions which anneal to complementary anti-codon sequences, associated with reactive units, thereby' ) bringing the reactive units together in a sequence-specific manner to'create a reaction product. Since nucleic acid hybridization is sequence 7 dpecific, the result of a nucleic acid-templated reaction is the translation'of a specific nucleic acid sequence into a corresponding reaction product. {01401 As shown in Figure 1, the ability of single-stranded nucleic acid templates to catalyze the sequence-specific oligomerization of complementary oligonucleotides has been demonstrated (Inoue et al. (1981) J. Am. CHEM. Soc. 103: 7666; Inoue et al. (1984) J. Mo. BioL. 178: 669-76). This discovery was soon followed by findings that DNA or RNA templates can catalyze the oligomerization of complementary DNA or RNA mono-, di-, tri-, or oligonucleotides (Inoue et al. (1981) 1. AM. CHEM. Soc. 103: 7666; Orgel et al. (1995) Acc. CHEM. REs: 28: 109-118; Rembold et al. (1994) J. MOL. EvoL. 38: 205; Rodriguez et at (1991) J. MoL. EVOL. 33: 477; Chen el al (1985) J. MOL. BIOL. 181: 271). DNA or RNA templates have since been shown to accelerate the formation of a variety of non-natural nucleic acid analogs, including peptide nucleic acids (Bohler et al. (1995) NATURE 376: 578), phosphorothioate- (Herrlein et at (1995) J. AM. CHEM. Soc. 117: 10151-10152), 5 phosphoroselenate- (Xu et al. (2000) J. AM. CHEM. Soc. 122: 9040-9041; Xu et at (2001) NAT. BIOTECHNOL. 19: 148-152) and phosphoramidate- (Luther et al. (1998) NATURE 396: 245-8) containing nucleic acids, non-ribose nucleic acids (Bolli et al. (1997) CHEM. BIoL. 4: 309-20), and DNA analogs in which a phosphate linkage has been replaced with an aminoethyl group (Gat el al. (1998) BIOPOLYMERS 48: 19-28). Nucleic acid templates can also catalyze amine ) acylation between nucleotide analogs (Bruick et a. (1996) CHEM. BIOL. 3: 49-56).

WO 2004/016767 PCT/US2003/025984 - 23 101411 Although nucleic acid templates haVe been, demonstrated to accelerate the formation of a variety of non-natutal nucleipciei analogues, nearly all'of these reactions Were designed to proceed through transition states closely resembling the natutAl nucleic acid backbone (Figure 1), typically affording prodiucts-that preserve the same six-bond backbone 5 spacing between nucleotide units. The motivation behind this design presuniably was the assumption that the rate enhancement provided by nucleic acid templates depends on precise alignment of reactive groups, and the precision of this alignment is maximized when the reactants and products mimic the structure of the DNA and RNA backbones. Evidence in support of the hypothesis that nucleic acid-templated synthesis can only generate products that 10 resemble the hucleic acid backbone comes from the well-nown difficulty of macrocyclization in organic synthesis (Illuminati et at (1981) AcC.-CHEM.'RES. 14: 95102; Woodward et aL (1981) J. AM. CHEM. Soc. 103: 3210-3213). The rate'enhancement of intramolecular ring closing reactions compared with their internolecular counterjiarts is known to diminish quickly as rotatable bonds are added between reactive groups, such that linking reactants witli a flexible 14 15 carbon linker hardly affords any rate acceleration (Illuminati et al (198 l supra). [01421 Because synthetic molecules of interest do not in general resemble nucleic acid backbones, the use of nucleic acid-templated synthesis to translate nucleic acid sequences into synthetic molecules is useful broadly only if synthetic molecules other than nucleic acids and nucleic acid analogs can be synthesized in a nucleic acid-templated fashion. Significantly, as 20 shown herein, nucleic acid-templated synthesis is indeed a general phenomenon and can be used for a variey of reactions and conditions to generate a diverse range of compounds, specifically including compounds that are not, and do not resemble, nucleic acids or nucleic acid analogs. More specifically, the present invention extends the ability to amplify and evolve libraries of chemical compounds beyond natural biopolymers. The ability to synthesize chemical 25 compounds of arbitrary structure allows researchers to write their own genetic codes incorporating a wide range of chemical functionality into novel backbone and side-chain structures, which pennits the development of novel catalysts, drugs, and polymers, to name a few examples. For example, the direct amplification and evolution of molecules by genetic selection permits the discovery of entirely new families of artificial catalysts which possess 30 activity, bioavailability, solvent, or thermal stability, or other physical properties (such as fluorescence, spin-labeling, or photolability) that may be difficult or impossible to achieve using the limited set of natural protein and nucleic acid building blocks. Similarly, developing WO 2004/016767 PCT/US2003/025984 -24 methods to amplify and directly evolve synthetic small molecules by iterated cycles of niutation and selection permits the isolation of novel ligands or drugs with properties superior to those isolated by traditional'rational design or combinatorial screening drug discovery methods. Additionally, applying this approach to the identification and development of polymers of 5 significance in material science can permit the evolution of new plastics or other polymers. [01431 In general, nucleic acid-templated synthesis as performed herein involves 1) providing one, or more nucleic acid templates bptionally associated with a reactive unit, and 2) contacting the one or more nucleic acid templates with one or more transfer units including an anti-codon associated with a reactive unit. The anti-codons of the transfer units are designed-to 10 hybridize to the nucleic acid template. In certaixi embodiments of the invention, the transfer unit comprises a single moiety 'simultaneously incorporating the hybridization capability of the anti codon unit and the chemical functionality of the'reaction unit. After the transfer units have hybridized to the nucleic acid template in a sequence-specific manner, the reactive units present on the transfer units and/or the nucleic acid template come into reactive proximity to react and 15 generate a reaction product. Preferably, the oligonucleotide portion of the transfer unit is removed once the reactive units have reacted to generate the reaction product or an intermediate of the reaction product. Significantly, the sequence of the nucleic acid template can later be determined, to permit decoding of the synthetic history of the attached reaction product and, thereby, its structure. This method may be used to synthesize one molecule at a time or may be 20 used to synthesize thousands to millions of compounds using combinatorial methods. 101441 In one embodiment, the template molecule optionally is associated with a reactive unit prior to interaction with any transfer units. Thus, as shown in Figure 2, the template can be connected by a covalent bond to a reactive unit, either directly or via a linker. Alternatively, the template can be connected by a noncovalent linkage. For example, the template can be 25 biotinylated, generally at a fixed location on the molecule, and can stably interact with a reactive unit associated with an avidin or streptavidin moiety. For ease of synthesis, the reactive unit is preferably placed at or near the 5' end of the template in some embodiments as shown in Figure 2. In other embodiments, placement of the reactive unit at an internal position of the template or at the 3' end is preferred. The template molecule also includes at least one codon capable of 30 annealing to an anti-codon of a transfer unit. During synthesis, the transfer unit anneals to the FLu I U,3hUUvr"~ WO 2004/016767 -25 conie unit into reactive roxitY with the reactive unit of the template to codon, bringing its reac-tive ntit eciepo produce a reaction product. [01451 in another embodiment, as shown in Figure 3, the template is not mitially e unit, but permits the nucleic acid-templated synthesis of at least two associated with a reactiveui, p- -- udsa es'w I~ -t TIi tmlemolecule includes es'w 5 reactive units disposed with two transfe units. - template d w in eat as t 5 reacive Uits . different codons, each capable of annealing to a differantai-codon disposed codons, anneach tals jo the corrcpondig codon in the template to bring' Trhe anti-codpn in each transfer unit anne t wit on ante topou' ' the reactive units of each transfer unit into reactive proxim y reaction product. 10 [01461 In another embodiment, as shown in Figure 4, the template can bring together, eitr sqentialY a plurlity of transfer units in a sequence-specific manner. either simultaneously or sequenialy, aPlrit . . dNihoeaohri The reactive units on each annealed transfer unit:can tlgen be reacted vith one another in a polymenzation process to produce a polymer. Using this approa s e tO generate a variety of noii-natural polymers. The polymerization maybe a step-by-step process or igiay be a ety rated in one reaction sequence.

15 simultaneous process whereby all the annealed monomers are reach I. TEMPLATE CONSIDERATIONS [01471 The nucleic acid template can direct a wide variety of chemical reactions without obvious structural requirements by sequence-specifically recruiting reactants linked to complementary oigonucleotides. As discussed, the nucleic acid mediated fonnat permits 20 reactions tat may not be possible using conventional synthetic approaches. During synthesis, the template hybridizes or anneals to one or more transfer units to direct the synthesis of a reaction product, which during certain steps oftemplated synthesis remain associated with the template. A reaction product then is selected or screened based on certain criteria, such as the ability to bind to a preselected target molecule. Once the reaction product has been identified, 25 the associated template can then be sequenced to decode the synthetic history of the reaction product. Furthermore, as will be discussed in more detail below, the template may be evolved to guide the synthesis of another chemical compound or library of chemical compounds. (i) Template Format [01481 The template may be based on a nucleic acid sequence, for example, a DNA, an 30 RNA, a hybrid of DNA and RNA, or a derivative of DNA and RNA, and may be single- or WO 2004/016767 PCT/US2003/025984 -26 double-stranded. The design of a particular template may vary depending upon the.type of rucleic acid templated synthesis contemplated. [01491 Figure 5 shows a variety of templates that may be useful in the practice of the invention. Figures 5A-C are schematic representations of templates including two codons for' 5 interaction with complementary anti-codons of two transfer units. These templates can be used in the type of nucleic acid-templated synthesis wherelno reactive units are linked to tle template at the initiation of synthesis; for example, when two transfer units anneal to the ternplate to bring , heir reactive units into reactive proximity to create a reaction product. One such example is polymerization. Nevertheless, the templates can be associated with a reactive unit prior to 10 annealing of the transfer units. Figures SD-F are schematic representations of templates that can be used in the type of nucleic acid-templated synthesis where one reactive unit is linked to the template at the initiation of synthesis, for example, when one transfer unit anneals to the template to bring its reactive unit into reactive proximity with the other reactive unit linked to the template to create'a reaction product. 15 [0150 Figure SA shows a template comprising in a 5' to 3' direction, a nucleotide sequence encoding a first primer binding site (PBS 1) or a sequence complementary thereto, a nucleotide sequence encoding a first codon (C1) that anneals to an anti-codon sequence of a first transfer unit, a nucleotide sequence encoding a second codon (C2) that anneals to an anti-cpdon sequence of a second, different transfer unit, and a nucleotide sequence encoding a second 20 primer binding site (PBS2) or a sequence complementary thereto. The primer binding sites, although optional, are preferred in some embodiments to facilitate PCR-based amplification of templates. As will be discussed in more detail below, the CI sequence is selected so as to minimize cross-reactivity with the anti-codon sequence of the second transfer unit, and the C2 sequence is selected so as to minimize cross-reactivity with the anti-codon sequence of the first 25 transfer unit. As shown in Figure 5A, the CI and C2 sequences are separated by one or more intervening bases. In other words, the C1 and C2 sequences do not directly abut one another. During nucleic acid templated synthesis, both the first and second transfer units are capable of binding to the template at the same time. [0151] Figure 5B shows a template similar to that shown in Figure 5A, except there are 30 no intervening bases disposed between CI and C2. In other words, the Cl and C2 sequences directly abut one another. As with the template of Figure 5A, during nucleic acid templated WO 2004/016767 PCT/US2003/025984 -27 synthesis, both the first and second transfer units ate capable to binding to the template at the same time. (0152] Figure 5C shows a template similar. to those shown in Figures 5A and 5B, except that the sequence of CII overlaps the sequence of C2. ,Unlike the templates of Figures SA and 5 5B, during nucleic acid templated synthesis, the fitst and second transfer units cannot bth bind to the template |at the same time. Thus, unless the template is associated with a reactive unit prior to the iitation of synthesis, a thirdcodhn',should normally be present, so that two reactive units can anneal simultaneously to the template to permit the reaction to proceed. This type of template can require a step-by-step approach to the synthesis of the reaction product. For 10 example, the transfer units with anti-codons to I are added first, allowed to hybridize and react, and then removed before the transfer units 'with anii-codons to C2 are added. [0153] Figures 5D-SF show templates similarfo the template shown in Figure 5A, except that the template also includes a reactive unit (R) associated with, for example, covalently linked to,'the template. It is understood, however, that the'templates shown in both Figure 5B 15 and Figure 5C may also comprise a reactive unit (R) associated with the corresponding I template, as shown in Figures 5D-5F. To the extent that a template is associated with a reactive unit, the nucleotide sequence of the template further comprises a sequence of nucleotides or sequence tag that uniquely identifies the reactive unit associated with the template. Following template mediated synthesis, the reactive unit actually attached to the template that participated 20 in the reaction to generate the reaction product may be identified by reading the sequence of the sequence tag. [0154] In Figure 5D, R is linked to the template at a location in the vicinity of the 5' terminal end, for example, at the 5' end of the template or downstream of the 5' end of the template. In Figure SE, R is linked to the template at a location between the 5' terminal end and 25 the 3' terminal end. In this particular case, R is located at a position between C1 and C2, and represents an example of the T type template architecture discussed in more detail below. In Figure 5F, R is linked to the template at a location in the vicinity of the 3' terminal end, for example, at the 3' end of the template or upstream of the 3' end of the template. 10155] It is contemplated that each of the templates shown in Figures 5A-F, may 30 comprise one or more restriction endonuclease sites. For example, with reference to Figure SA, the template may comprise a restriction endonuclease site disposed between (i) PBSI and C1, (ii) WO 2004/016767 PCT/US20031025984 -28 C1 and C2, and (iii) C2 and PBS2. The restriction endonuclease sites facilitate the use of nucleic acid cassettes to easily introduce various'sequences to replace. the PB SI sequence, the C1, sequence, the C2 sequence, the PBS2 secluence, or any combination thereof. 101561 In addition, the template may also incorporte a hairpin loop on one end 5 terminating in a reactive unit that can interact with one or more reactive units associated with transfer units. For example; a DNA template can comprise a hairpin loop terminating in a 5' amino group, which may or may not be protected. The amino group may act as An initiation, point for formation of an unnatural polymer, or may belmodified to bind a small molecule scaffold for subsequent modification by reactive units of other transfer units. 10 [0157] The length of the template may vary greatly depending upon the type of the nucleic acid-templated synthesis contemplated. For example, in certain embodiments, the template may be from 10 to 10,000 nucleotides in length, from 20 to 1,000 nucleotides in length, from 20 to 400 nucleotides in length, from 40 to 1,000 nucleotides in length, or from 40 to 400 nucleotides in length. The length of the template will of course depend on, for example, the 15 length of the codons, the complexity of the library, the complexity and/or size of a reaction product, the use of spacer sequences, etc. (H) Codon Usage [0158] It is contemplated that the sequence of the template may be designed in a number of ways without going beyond the scope of the present invention. For example, the length of the 20 codon must be determined and the codon sequences must be set. If a codon length of two is used, then using the four naturally occurring bases only 16 possible combinations are available to be used in encoding the library. If the length of the codon is increased to three (the number Nature uses in encoding proteins), the number of possible combinations increases to 64. If the length of the codon is increased to four, the number of possible combinations increases to 256, 25 Other factors to be considered in determining the length of the codon are mismatching, frame shifting, complexity of library, etc. As the length of the codon is increased up to a certain point the number of mismatches is decreased; however, excessively long codons likely will hybridize despite mismatched base pairs. [0159] Although the length of the codons may vary, the codons may range from 2 to 50 30 nucleotides, from 2 to 40 nucleotides, from 2 to 30 nucleotides, from 2 to 20 nucleotides, from 2 to 15 nucleotides, from 2 to 10 nucleotides, from 3 to 50 nucleotides, from 3 to 40 nucleotides, WO 2004/016767 PCT/US2003/025984 -29 from 3 to 30 nucleotides, from 3 to 20 nucleotides, from 3 to. 15-nucleofides, from 3-to 10 nacleotides, from 4 to 50 nucleotides, frdm 4 to 40 nucleotides, from 4 to 30 nucleotides, from 4 to 20 nucleotides, from 4 to 15 nucleotides, from 4 to 10 Au.leotides, from 5 to 50 nucleotides, from 5 to 40 nucleotides, from 5 to 30 nucleotides, from 5 to 20 nucleotides, from 5 to 15 5 nucleotides, from 5 to 10 nucleotides, from 6 to 50 nucleotides, from 6 to 40 nucleotides; from 6 to 30 nucleotides, from 6 to 20 nucleotides, from 6 to 15 nucleotides, from 6 to 10 nucleotides, from 7 to 50 nucleotides, from 7 to 40 nucleotides, from 7 to 30 nucleotides, from 7 to, 20 inucleotides, from 7 to 15 nucleotides, from 7 to 10 nucleotides, from 8 to 50 nuckotides, from 8 to 40 nucleotides, from 8 to 30 nucleotides, from 8 to 20 nucleotides, from 8 to 15 nucleotides, 10 from 8 to 10 nucleotides, from 9 to 50 nucleotides, from 9 to 40 nucleotides, from 9 to 30 nucleotides, from 9 to 20 nucleotides, from 9 to 15 nucleotides, from 9 to 10 nucleotides. Codons, however, preferably are 3, 4, 5, 6, 7, 8, 9 or 10 nucleotides in length. [0160] In one embodiment, the set of coons used in the template maximizes the number of mismatches between any two codons within a codon set to ensure that only the proper anti 15 codons of the transfer units anneal to the codon sites of the template. Furthermore, it is important that the template has mismatches between all the members of one codon set and all the codons of a different codon set to ensure that the anti-codons do not inadvertently bind to the wrong codon set. For example, with regard to the choice of codons n bases in length, each of the codons within a particular codon set (for example, C1 in Figure 5A) should differ with one 20 another by k mismatches, and all of the codons in one codon set (for example, C1 in Figure 5A) should differ by m mismatches with all of the codons in the other codon set (for example, C2 in Figure 5A). Exemplary values for n, k, and m, for a variety of codon sets suitable for use on a template are summarized in Table 1. TABLE I 2 1 1 3 1 1 3 2 1 3 2 2 4 1 1 4 2 1 WO 20041016767 PCT/US2003/025984 -30 4 2 2 4 3 1 4 3 2 4 -3 3 5 1 1 5 2 ,i'l . 5 2 2 5 3 5 ,3 2 5, 3 3' -5 4 1 5 4 2 5 4 3 5 4 4 6 1 1 6 2 1 6 2 2 6 3' 1 6 3 2 6 3 3 6 4 1 6 4 2 6 4 3 6 4 4 6 5 1 6 5 2 6 5 3 6 5 4 6 5 5 7 1 1 7 2 1 WO 2004/016767 PCT/US2003/025984 -31 7 2 .2 7 3 1 7 3 2 7 3 3 7 4 1 7 4 2 7 4 3 7 5 1 7 5 2 7 5. 3 7 5 4 7 5 5 7 6 1 7 6 2 7 6 3 7 6 4 7 6 5 7 6 6 8 1 1 8 2 1 8 2 2 8 3 1 8 3 2 8 3 3 8 4 1 8 42 8 4 3 8 4 4 8 5 1 8 5 2 WIU 2004/016767 PCT/US2003/025984 -32 '8 5 3 8 5 4 8 5 5 8 6 1, 8 "6 2 8 6 '3 8 6 4 8 6 5 8 ,6 6 8, 7 1 8 7* 2 8 3 8 7 4 8 7 5 8 7 6 8 7' 7 9 1 1 9 2 1 9 2 2 9 3 1 9 3 2 9 3 3 9 4 1 9 4 2 9 4 3 9 4 4 9 5 1 9 5 2 9 5 3 9 5 4 9 5 5 WO 2004/016767 PCT/US2003/025984 - 33 9 6 1 9 6 2 9 6 4 9 6 4 9 6 6 966 9 7 1 9 7 2 9 7 3 9 7 4 9 7 5 9 7 6 9 7 7 9 8 1 9 8 2 9 8 3 9 8 4 9 8 5 9 8 6 9 8 7 9 8 8 10 1 1 10 2 1 10 2 2 10 3 1 10 3 2 10 3 3 10 4 1 10 4 2 10 4 3 10 4 4 WO 2004/016767 PCT/US2003/025984 -34 5 1 1W 5 2 10 5 3 10 5 4 .10 5 5 10,6 1 .10 6 2 10 6 3 .o 6 10 ,6 4 10 6 5' 10. 6 6 10 7: 1 10 7 2 10 7 3 10 7 4 10 7 5 10 7 6 10 7' 7 10 8 1 10 8 2 10 8 3 10 8 4 10 8 5 10 8 6 10 8 7 10 8 8 10 9 1 10 9 2 10 9 3 10 9 4 10 9 5 WO 2004/016767 PCT/IUS2003/025984 I -35I 10 9 6 10 9 7 10 9' 8 .10 9 9 rTt 11 1 1, *11 2 '2 11 3. 3 11 3- 2 11.3 .3' 11 ' 4 *1 11 '4- 2' 11 4 3 11 4 4 11 5. 11 5 ' 2 11 5 3 11 5 4 11 5 5 11 6 1 11 6 2 11 6 3 11 6 4 11 6 5 11 6 6 11 7 1 11 7 2 11 7 3 11 7 4 11 7 5 11 7 6 WO 2004/016767 PCT/US2003/025984 -36 11 7 7 11 8 1 11 8 3 11. 8 4. 11 8 5 11 8 6 11 8 7 T1 8 8 11 9 1 11 9 2 11 9 3 11 9 4 11 9 5 11 9 6 11 9 7 11 9 8 11 9 9 11 10 1 11 10 2 11 1 11 10 4 11 10 5 11 10 6 11 10 7 11 10 8 I1 10 9 11 10 10 12 1 1 12 21 12 2 2 WO 2004/016767 PCT/US2003/025984 -37 n k n~ 12 3 1 12 3 2 12 3 3 12 4 1 12 4 2 12-4 3 .12 44 12 5 ' 12 ,5 2 12 5 3 12 5 . 4 12 5 5, 12. 6 1 12 6 2 12 6 3 12 6 4 12 6 5 12 61 6 12 7 1 12 7 2 12 7 3 12 7 4 12 7 5 12 7 6 12 7 7 12 8 1 12 8 2 12 8 3 12 8 4 12 8 5 12 8 6 WO 2004/016767 -38 12 8 7 12 8 8 12 9 1 12 9 2 12. 9 3 12 9 A 12 9 5 12 9 6 12 9 7 12 9 -8 12 9 9 12 10 1 12 10 2 12 10 3 12 10 4 12 10 5 12 10 6 12 10 7 12 10 8 12 10 9 12 10 10 12 11 1 12 11 2 12 11 3 12 11 4 12 11 5 12 11 6 12 11 7 12 11 8 12 11 9 12 11 10 WO 2004/016767 PCT/US2003/025984 -1 -39 '12 1 1 13 1 1 13 2 1 13 2 2. .13 3 1 13 '3' 2 . 13 '3 13 4 13 4. 2 13, 4 .3 1314 ' 4 13 5 1', 13 5 2 13 5 3 13 5 4 13 5 ' 5 13 6 1 13 6' 2 13 6 3 13 6 4 13 6 5 13 6 6 13 7 1 13 7 2 13 7 3 13 7 4 13 7 5 13 7 6 13 7 7 13 8 1 13 8 2 WO 2004/016767 PCTfUS20031025984 -40 13, 8 3 13 8 4 13 8 5 13 8 6 13,8 7 13 8 8 13 9 1 13 9 2 13 9 3 13 9 4 13 9 5 13 9 '6 13 9 7 13 9 8 13 9 9 13 10 1 13 10 2 13 10 3 13 10 4 13 10 5 13 10 6 13 10 7 13 10 8 13 10 9 13 10 10 13 11 1 13 11 2 13 11 3 13 11 4 13 11 5 13 11 6 WO 2004/016767 PCT/US2003/025984 -41 13 11 .7 11 8 13 11 9 13 11 10 13 , 11 13 12 1 13 12 2. 13 12 3 13 12 4 13 12 5 13 12 6 13 12 7 13 12 8 13 12 9 13 12 10 13 12 11 13 12 12 14 1 1 14 2 1 14 2 2 14 3 1 14 3 2 14 3 3 14 4 1 14 4 2 14 4 3 14 4 4 14 5 1 14 5 2 14 5 3 14 5 4 WO 2004/016767 PCT[US2003/025984 -42 '14 5 5 ' 146 1 14 6' 2 14 6 i3 14 6 4 14 -6' 52 14 6 6 14-166 14 7' 1. 14 7 2 14, 7 3 14-7 4, 14 '7 51 147 6 14 7 7 14 8 1 14 8 2 14 8 3 14 8 4 14 8 5 14 8 6 14 8 7 14 8 8 14 9 1 14 9 2 14 9 3 14 9 4 14 9 5 14 9 6 14 9 7 14 9 8 14 9 9 WO 2004/016767 PCT/US2003/025984 -43 14 10 1 14 10 2 14 10 3 14 10 4 14 10 5 . 14 10 6 '14 10 '7 14 10 8 14 10 9 14 10 10 14 11 1 14 11 2 14 11 3 14 11 4 14 11 5 14 11 6 14 11 7 14 11 8 14 11 9 14 11 10 14 11 11 14 12 1 14 12 2 14 12 3 14 12 4 14 12 5 14 12 6 14 12 7 14 12 8 14 12 9 WO 2004/016767 PCT/US20031025984 44 n k m 14.- 1.2 1U 14 12 '12 14 13 1 S 14--13 2. 14.1 13 .3 1-4, 13 .4)' 14 13 5 14 13 14 13 7 14 13, 8 14 .13 9 14 13 10 14 13 11 14 13 12 14 13 13 15 1 1 15 2 1 15 2 2 15 3 1 15 3 2 15 3 3 15 4 1 15 4 2 15 4 3 15 4 4 15 5 1 15 5 2 15 5 3 15 5 4 15 5 5 15 6 1 WO 2004/016767 PCTf/US2003/025984 -45 15 6 2 15 6 3 15 6 4 15 6 5 15 6 6 15 7 1 15 7 2, 15 7 3 T5 7 4 15 7 .5 15 7 6 15 7 7 15 8 1 15 8 2 IS 8 3 15 8 4 15 8 5 15 8 6 15 8 7 15 8 8 15 9 1 15 9 2 15 9 3 15 9 4 15 9 5 15 9 6 15 9 7 15 9 8 15 9 9 15 10 1 15 10 2 WO 2004/016767 PCT/US2003/025984 -46' 15 10 3 15 10 .4 15 10 5 1510 6 15 10 7 15 10 8 15 10 10 15 ii 1 15 11, 2' 15' 11 3 15'T1174 15, 11 5 15 11 6 15 11 7 15 11 8 15 11 9 15 1I 10 15 11 11 15 12 1 15 12 2 15 12 3 15 12 4 15 12 5 15 12 6 15 12 7 15 12 8 15 12 9 15 12 10 15 12 11 15 12 12 WO 2004/016767 PCT/US2003/025984 -47 15 13 1 15 13 2 15 13 3 15 13 4 15 13 5 15 13 6 15 13 7, 15 13 8 1'5 13 :9 15 13 10 15 13 11 15 13 12 15 13 13 15 14 1 15 14 2 15 14 3 15 14 4 15 14 5 15 14 6 15 14 17 15 14 8 15--14 9 15 14 10 15 14 11 15 14 12 15 14 13 15 14 14 [01611 Using an appropriate algorithm, it is possible to generate sets of codons that maximize mismatches between any two codons within the same set, where the codons are n bases long having at least k mismatches between any two codons. Since between any two WO 2004/016767 PCT/US20031025984 , -48j codons, there must be at least k mismatches, any tjo subcodons of n - (k 1) bases must have at least one mismatch- This sets an uyper limit .f 4"r on' the size of any (n, k) codon set Such an algorithm preferabij starts with the 4" 1 possible subcodons of length n (k 1) and then tests all combinations of adding k - 1 bases for those'that always maintain k mismatches. All possible 5 (n, k) sets 6an be generated.for n < 6.' For n >'6, the 4 "t1 upper limits of coddns cannot be met and a "full" packing of viable codons is mathematically impossible. In addition to there, being at least one mismatch k between codons within'thp'same codon set, there should also be at least' one mismatch m between all. the codons of one codon set and all the codons of another codon set. Using this approach, different sets of codons cai Pe generated so that no codons are repeated. 10 101621 By way of example, four (n=5, k3, m=l) sets, each with 64 codons, can be chosen that always have at least one mismatch between any two codons in different sets and at least three mismatches between codons in the same set.' TABLE 2: Sequences of (5,3,1) Codon Set I CCCTC CCGAG CCTCT CCAGA CGCGT CGGCA CGTAC CGATG CTCCG , CTGC CTTTA CTAAT CACAA CAGTT CATGG CAACC GCCCA GCGGT GCTTG GCAAC GGCAG GGGTC GGTGA GGACT GTCTT GTGAA GTTCC GTAGG GACGC GAGCG GATAT GAATA TCCGG TCGCC TCTAA TCATT TGCTA TGGAT TGTCG TGAGC TTCAC TTGTG TTTGT TTACA TACCT TAGGA TATTC TAAAG ACCAT ACGTA ACTGC ACACG AGCCC AGGGG AGTTT AGAAA ATCGA ATGCT ATTAG ATATC AACTG AAGAC ATCA AAAGT I 15 TABLE 3: Sequences of (5,3,1) Codon Set 2 CCCAC CCGTG CCTGT CCACA CGCCT CGGGA CGTTC CGAAG CTCGG CTGCC CTTAA CTATT CACTA CAGAT CATCG CAAGC GCCGA GCGCT GCTAG GCATC GGCTG GGGAC GGTCA GGAGT WO 2004/016767 PCT/US2003/025984 -49 GTCAT GTGTA GTTGC GTACG GACCC GAGGG GATTT GAAAA TCCCG TCGGC TCTTA TCAAT TGCAA TGGTT TGTGG TGACC TTCTC TTGAG TTTCT TTAGA TACGT TAGCA TATAC TAATG ACCTT ACGAA ACTCC ACAGG AGCGC AGGCG AGTAT AGATA ATCCA' ATGGT ATTTG ATAAC AACAG AAGTC AATGA AAACT TABLE 4: Sequences of (5,3,1) Codon Set 3 Codon Seq' CoonsC Codon e d Seq CoddnSeq Codoin CCCTG CCGAC CCTCA CCAGT CGCAT CGGTA CGTGC CGACG CTCCC CTGGG CTTTT CTAAA CACGA CAGCT CATAG CAATC- GCCAA GCGTT GCTGG GCACC GGCTC GGGAG GGTCT GGAGA GTCGT GTGCA GTTAC GTATG GACCG GAGGC GATTA GAAAT TCCGC TCGCG TCTAT TCATA TGCCA TGGGT TGTTG TGAAC TTCAG TTGTC TTTGA TTACT TACTT TAGAA TATCC TAAGG ACCCT ACGGA ACTTC ACAAG AGCGG AGGCC AGTAA AGATT ATCTA ATGAT ATTCG ATAGC AACAC AAGTG AATGT AAACA TABLE 5: Sequences of (5,3,1) Codon Set 4 CCCAG CCGTC CCTGA CCACT CGCTT CGGAA CGTCC CGAGG CTCGC CTGCG CTAT CTATA CACCA CAGGT CATTG CAAAC GCCTA CGAT GCTCG GCAGC GGCAC GGGTG GGTGT GGACA GTCCT GTGGA GTTTC GTAAG GACGG GAGCC GATAA GAATT TCCCC TCGGG TCTTT TCAAA TGCGA TGGCT TGTAG TGATC TTCTG TTGAC TTTCA TTAGT TACAT TAGTA TATGC TAACG ACCGT ACGCA ACTAC ACATG AGCCG AGGGC AGTTA AGAAT ATCAA ATGTT ATTGG ATACC WO 2004/016767 PCT/US20031025984 -50 AACTC AAGAG , AATCT AAAGA . 10163] Similarly, four (n-=6, k=4, m=2) sets as,shown below, each with 64 codons, can be chosen that always have at least two misinatclies between any two codons in different codon sets and at least four'nismatches, between coaons 4 n the same cbdon set. 5 TABLE 6: Sequences of (6,4,2)- Codon Set 1 Co 6don Seq. ;Cdoh 0&q CooCez d2dfSj ogtSq - oo\e CCCTCC 'TCGAAC, CCGCTG TCTCCA CGGTAT TCATTT CCAGAA TGCACT, CGCCGA TGGGTA CTCAAG TGTTGC CGTGCG TGACAG CGAATC, TTCCTC CTACCT TTGTCG CTGGGC' TTTGAT CTTTTA TTAAGA CATCAC TACTAA CACGTT -TAGCGT CAGACA TATATG GCGGCT TAAGCC CAATGG ACCCAT GCCATA ACGTGA GGCGAC ACTGTC GCTTAG ACAACG GCACGC AGCTTG GGATCA AGGCCC GGGAGG AGTAAA GGTCTT AGAGGT GTTACC ATCGCA GTCTGT ATGATT GTGCAA ATTCGG GAGTTC ATATAC GTAGTG AACAGC GACCCG AAGGAG TCCGGG AATTCT GATGGA AAACTA GAAAAT CCTAGT TABLE 7: Sequences of (6,4,2) Codon Set 2 CodoSn, 6 .iC'a - q CCCCTC TCGGGC CCGTCG TCTTTA CGGCGT TCACCT CCAAGA TGCGTT CGCTAA TGGACA CTCGGG TGTCAC CGTATG TGATGG CGAGCC TTCTCC CTATTT TTGCTG CTGAAC TTTAGT CTTCCA TTAGAA CATTGC TACCGA CACACT TAGTAT CAGGTA TATGCG GCGATT TAAATC CAACAG ACCTGT GCCGCA ACGCAA GGCAGC ACTACC GCTCGG ACAGTG GCATAC AGCCCG GGACTA AGGTTC GGGGAG AGTGGA GGTTCT AGAAAT GTTGTC ATCATA GTCCAT ATGGCT GTGTGA ATTTAG GAGCCC ATACGC WO 20041016767 PCT/US2003/025984 - 51 GTAACG. AACGAC GACTTG AAGAGG .TCCAAG AATCTT GATAAA AAATCA GAAGGT CCTGAT TABLE 8: Sequences of (6,4,2) Codon Set 3 odon Seq. Cpdon Seq. Codon Seq. CodoiSeq. Codon Seq. Codon Seq CCCGAC- TCGCCC CCGAGG TCTAAA CGGGCT TCAGGT CCATCA TGCCAT CGCATA TGGTGA CTCCCG TGTGTC CGTTAG TGAACG CGACGC TTCAGC CTAAAT TTGGAG CTGTTC TTTTCT CTTGGA. TTACTA CATACC TACGCA CACTGT TAGATT CAGCAA TATCGG GCGTAT TATAC CAAGTG ACCACT GCCCGA ACGGTA GGCTCC ACTTGC GCTGCG ACACAG GCAATC AGCGGG. GGAGAA AGGAAC GGGCTG AGTCCA GGTAGT AGATTT GTTCAC ATCTAA GTCGTT ATGCGT GTGACA ATTATG GAGGGC ATAGCC GTATGG AACCTC GACAAG AAGTCG TCCTTG AATGAT GATTTA AAAAGA GAACCT CCTCTT TABLE 9: Sequences of (6,4,2) Codon Set 4 onSe Codon Seq CodonSdq ConSq Cdon Seq. Coddn~q CCd. 'd CCCAGC TCGTTC CCGGAG TCTGGA CGGATT TCAAAT CCACTA TGCTGT CGCGCA TGGCAA CTCTTG TGTACC CGTCGG TGAGTG CGATAC TTCGAC CTAGGT TTGAGG CTGCCC TTTCTT CTTAAA TTATCA CATGTC TACATA CACCAT TAGGCT CAGTGA TATTAG GCGCGT TAACGC CAAACG ACCGTT GCCTAA ACGACA GGCCTC ACTCAC GCTATG ACATGG GCAGCC AGCAAG GGAAGA AGGGGC GGGTCG AGTTTA GGTGAT AGACCT GTTTGC ATCCGA GTCACT ATGTAT GTGGTA ATTGCG GAGAAC ATAATC GTACAG AACTCC GACGGG AAGCTG TCCCCG AATAGT GATCCA AAAGAA GAATTT CCTTCT 5 WO 2004/016767 PCT/US2003/025984 -52-a [0164] Codons can also be chosen to increase control over the GC content and, therefore, the melting temperature of the codon add anti-codon. Cbdons sets with a wide range in GC content versus -ATcbntent may result in reagents that'anneal with differeniefficiencies due to different melting temperatures. By screening fo' GC content among different (n, k) sets, the GC 5 content foi the codon sets can be optimized. For example, the four (6, 4, 2) c6don sets set forth in Tables 6-9 each contain 40 codons with identical .GC content (i.e., 50% Gp content) By using only these 40 codons at each position, all the reagents in theory will have comparable melting temperatures, removing potential biases in annaling that might otherwise affectlibrary. synthesis. Longer codons that maintain a large hpmber of mismatches such as those appropriate 10 for certain applications such as the reaction discovery system can also be chosen using this approach. For example,'by combining two (6, 4) sets together while matching low GC to high GC codons, (12, 8) sets with 64 codons all With 50% GC content can be generated for use in reaction discovery selections as well as other application where multiple mismatches might be advantageous. These codons satisfy the requirements for encoding a 30 xs- 30 matrix of 15 functional group combinations for reaction discovery. [0165] Although an anti-codon is intended to bind only to a codon, as shown in Figure 6A, an anti-codon may also bind to an unintended sequence on a template if complementary sequence is present. Thus, an anti-codon may inadvertently bind to a non-codon sequence as shown in Figure 6B. Alternatively, as shown in Figures 6C and 6D, an anti-codon might 20 inadvertently bind out-of-frame by annealing in part to one codon and in part to another codon (Figure 6C) or to a non-codon sequence (Figure 6D). Finally, as shown in Figure 6E, an anti codon might bind in-frame to an incorrect codon, an issue addressed by the codon sets described above by requiring at least one base difference distinguishing each codon. In Nature, the problems of noncoding sequences and out-of-frame binding (Figures 6B-D) are avoided by the 25 ribosome. The nucleic acid-templated methods described herein, however, do not take advantage of the ribosome's fidelity. Therefore, in order to avoid erroneous annealing as in Figures 6B-D, the templates can be designed such that sequences complementary to anti-codons are found exclusively at in-frame codon positions. For example, codons can be designed to begin, or end, with a particular base (e.g., "G"). If that base is omitted from all other positions in 30 the template (i.e., all other positions are restricted to T, C, and A), only perfect codon sequences in the template will be at the in-frame codon sequences. Similarly, the codon may be designed WO 2004/016767 PCTIUS2003/025984 -53 to be sufficiently long such that its sequence is unique and does not appear' elsewhere in a template. [01661 When the nucleic 'acid-templated synthesis-is used to produce a polymer, spacer sequences may also be placed between the codons to prevent frame shifting. More preferably, the bases of the template that encode each polymer subunit (the "genetic code" for the polymer) may be chosen hiorn Table 10 to preclude or minimize the possibility of out-6f-frame annealing. These genetip codes reduce undesired frameshifted nucleic acid-templated polymer translation and differ in the range of expected melting temperatures and in the minimum number of mismatches that result during out-of-frame annealing. 7 TABLE -1 0 Representative Genetic Codes for Nucleic Acid-templated Polymers That Preclude Out-Of-Frame Annealing I ~~ience P Nzme bf sbeC~os VVNT 36 possible codons NVVT 36 possible codons SSWT 8 possible codons SSST 8 possible codons SSNT 16 possible codons VNVNT or NVNVT 144 possible codons SSSWT or SSWST 16 possible codons SNSNT or NSNST 64 possible codons SSNWT or SWNST 32 possible codons WSNST or NSWST 32 possible codons where,V = A, C, orG, S=CorG, W=AorT, andN=A, C, G, orT 101671 As in Nature, start and stop codons are useful, particularly in the context of 5 polymer synthesis, to restrict erroneous anti-codon annealing to non-codons and to prevent excessive extension of a growing polymer. For example, a start codon can anneal to a transfer WO 2004/016767 PCT/US2003/025984 -54 unit bearing a small molecule scaffold or a start monomer unit for use in polymer. synthesis; the start monomer unit can be masked by a photolabile protecting group as shown in Example 5A. A stop codon, if used to terminate polymer synthesis, shouId not conflict with any other todons used in the synthesis and should be of the same general format as the other codons. Generally, a 5 stop codon, can encode a monomer unit that terminates polymerization by not providing a reactive group for further attachment. For example, a stop, monomer unit may contain a blocked reactive group such as an acetamide rather than a primary-amine as shown in Example 9A: In other embodiments, the stop monomer unit can include a biotinylated terminus that terminates the polymerization and facilitates purification of the resulting polymer. 10 (iW) Template Architecture 101681 As discussed previously, depending upon the type of nucleic acid-templated synthesis contemplated, the template may be further associated (for example, covalently coupled) with a particular reactive unit. Various templates useful in nucleic acid-templated synthesis are shown in Figures 7A-7G, and include templates referred to as the "end-6f helix" or 15 "E" templates (see, Figure 7A-C), "Hairpin" or "H" templates (see, Figure 7D), "Omega" or "D" templates (see, Figure 7E-F), or "T" templates (see, Figure 7G). 101691 Figures 7A-C show E type template architectures where the reactive units on the annealed templates (denoted by A) and transfer units (denoted by B) are separated by 1 base (Figure 7A), 10 bases (Figure 7B) and 20 bases (Figure 7C). Figure 7D, shows a H type 20 template architecture where the reactive unit is attached to the template (denoted by A) and the template folds back on itself to create a hairpin loop stabilized by a plurality of intramolecular bonds. As shown, the reactive units on the annealed template (denoted by A) and the transfer unit (denoted by B) are separated by I base. Figures 7E-F show omega type template architecture where the codon for the transfer unit, bearing reactive unit B, is separated from 25 reactive unit A on the template by 10 intervening template bases (Figure 7E) or by 20 bases (Figure 7F). In Figure 7E, the omega template comprises a three base constant region (0-3) and creates a seven base loop when the transfer unit anneals to the template. In Figure 7F, the omega template includes a five base constant region (0-5) and creates a fifteen base loop when the transfer unit anneals to the template. The loop gets larger as transfer units anneal to codons 30 further away from the constant region of the template. Figure 7G shows a T-type template architecture where the reactive units on the annealed template (denoted by A) and the transfer WO 2004/016767 PCT/US2003/025984 -55-, unit (denoted by B) are separated by 1 base: In Figure 1G; reactive unit A is attached at a location intermediate the 5' and 3' terminal ends of the template. Using this architecture, it is contemplated that th' reactive unit may be atached td the template at a loohtion at least 10, 20, 30, 40,' 50, 60, 70 bases or more downstream of the 5' end of the template and/or at least 10, 20, 5 30, 40, 50,'60, 70 bages or more upstream of the 3, end of the template.

(0170] The ability of the E type template ardhitectdre and the H type template architecture to facilitate nucleic acid mediated chemical syntheses is described in detail in Example 1. However, as a resultof perfofining nucleic acid mediated syntheses, it has leen discovered that certain reactions, referred to as distance dependent reactions, do not proceed 10 efficiently When the annealed reactive units on the template and transfer unit are separated by even small numbers of bases. Using the E ad- H type templates, certain distance- dependent reactions may only be encoded by template bases at the reactive end of the template, The new Q type template 6yercomes the distance dependence problems that can be experienced with the E and H type templates (see, Example 5). Furihermore, it has been discovered that the presence of 15 double-stranded nucleic acids between annealed reactive units can greatly reduce the efficiency of templated reactions because the flexibility of a single-stranded template is required. This may hinder performing two or more reactions in a single nucleic acid templated step using the E or H architectures even though the template may contain enough bases to encode multiple reactions. The new T type template overcomes this problem that can be experienced with the E and H type 20 templates (see, Example 5). 9 Templates [01711 The omega architecture permits distance dependent reactions to be directed efficiently by nucleotide bases far away from the reaction end of the template, effectively overcoming their distance dependence. By way of example, in the omega architecture, five 25 bases of the template are held constant at the 5'-end of the template (see, Figure 7F). The transfer units contain at their 3'-ends the complementary five bases but otherwise possess sequences that complement distal coding regions of the template. This permits the transfer unit to anneal to the distal coding regions of the template while still placing the reactive group of the transfer unit in close proximity by looping out large numbers of template bases that would 30 ordinarily prevent a distance dependent reaction from proceeding. The omega architecture WO 2004/016767 PCT/US2003/025984 -56 retains sequence specificity because the five bases of the transfer unit that complement the end of the template are insufficient by themselves to anneal to the template at room temperature. {01721 , The usefulness of this type of template architecture is apparent, for example, in nucleic acid-templated reductive amination reactions. These reactions are strongly distance 5 dependent and very little product is produced when the reaction is attempted using the hairpin or end-of-helix architectures with more than one base of distance between the annealed amine and aldehyde groups. In contrast, product forms efficiently using the omega architecture even when a region of the template 20 bases away from the reactive end is used to recruit thereagent (see, Example 5). No product is observed when the coding region of the transfer unit is mismatched, 0 despite the presence of five bases at the end of the transfer unit that are complementary to the end-of the template. 101731 By enabling distance-dependent nucleic acid mediated reactions to be encoded by bases far away from the reactive end of the template, the omega architecture expands the types of reactions that can be encoded anywhere on the template. 5 T Templates 101741 , The T architecture permits a single template to encode two distance-dependent reactions and in addition permits a template to undergo two different nucleotide-templated reactions in a single solution or in "one-pot." Using this architecture, the template can present a molecular scaffold through the non-Watson-Crick face of a base located in the center, rather than -0 the end, of the template (see, Figure 7G). This permits two transfer units to anneal to either side of the reactive unit attached to the template and react either simultaneously or in successive steps to give the product of two nucleotide-templated transformations. As expected, distance dependent reactions tolerate this architecture when reactive groups are proximal. Thus, the T type architecture permits two sequence-specific nucleic acid-templated reactions to take place on 25 one template in one solution, i.e., in one step. In addition to reducing the number of separate DNA-templated steps needed to synthesize a target structure, this architecture may permit three or more component reactions commonly used to build structural complexity in synthetic libraries. 10175] The omega and T architectures permit a broader range of template mediated 30 reactions that can be performed in fewer steps with other template architectures and are especially useful in distance-dependent reactions. The variety of available architectures provide WO 2004/016767 PCT/US2003/025984 -57 significant flexibility in the placement of reactive units on templates, particularly for the synthesis of 'small molecules. It is dontoinplated that the reactive unit including, for example, molecular scaffold'naybe associatedIrith.a template-at any site'along the template including the, 5'-end (e.g., end-of-helix architecture, omega architecture), the 3'-end (e.g., end-of-helix 5 architecture, omega architecture), at the end of a hairpin, loop (e.g., hairpin architecture), or in the middle of the template (e.g., T architecture). Preferably, the molecular scaff o ld is attached covalently to the template. However, in certain hembodinients, the molecular scaffold, like thb other reactive units; can be brought to'the template using a transfer unit, in which case the molecular scaffold is only associated with the tenpate through a non-covalent (here, hydrogen 10 bonding) interhetion. It is contemplated, however, that under certain circumstances it may be advantageous to covalently link the molecular scaffold or another reactive unit to the template to produce a T- or E-type template architecture: For reactions that are not distance dependent, the position of the Tnolecular scaffold along the template is more flexible because the reactive units brought to the template by the transfer units are able to react with the scaffold even if the 15 scaffold and reactive group are separated by many bases. (iv) Template Synthesis [0176] The templates may be synthesized using methodologies well known in the art. For example, the nucleic acid sequence may be prepared using any method known in the art to prepare nucleic acid sequences. These methods include both in vivo and in vitro methods 20 including PCR, plasmid preparation, endonuclease digestion, solid phase synthesis (for example, using an automated synthesizer), in vitro transcription, strand separation, etc. Following synthesis, the template, when desired may be associated (for example, covalently or non covalently coupled) with a reactive unit of interest using standard coupling chemistries known in the art. 25 101771 By way of example, it is possible to create a library of templates via a one-pot modular ligation reaction using oligonucleotide cassettes shown as discussed, for example, in Example 9C. Specifically, it is possible to combine short oligonucleotides representing all transfer unit annealing regions together with T4 DNA ligase in a single solution. Due to the sequence design of the oligonucleotide termini, the desired assembled template library is the 30 only possible product when the ligation is complete. This strategy requires 2n x m short oligonucleotides to assemble a library of n' templates, where n refers to the number of different WO 2004/016767 PCT/US2003/025984 -58 sequences per codon position and m refers to the number of codons per library member. Thus, for a two-codon template with 64 possible sequences per codon, 2 x 64 x 2 (256) oligonucleotides are required to assemble a library of 64 (4096) templates. The one-pot assembly of the templates for the 83-membered macrocyclic fimaramide library is discussed in 5 Example 9B. Excellent yields of the desired template library resulted from a 4 hour ligation reaction. Following ligation, T7 exonuclease was added to degrade the non-coding template strand (the desired coding strand is protected by its non-natural 5'-aminoethylene glycol linker),. This procedure can provide 20 nmoles of the 5' functionalized single-stranded teinplate library (sufficient material for thousands of DNA-templated library syntheses and selections) in about 6 10 hours. The constant 10-base primer binding regions at the ends of each template were sufficient to permit PCR amplification of as few as 1,000 molecules (10-21 mol) of template from this assembled material. 10178 Another approach for synthesizing templates is shown in Figure 8. In particular, Figure 8 shows a protocol for producing a template containing in a 5' to 3' direction,. a small 15 molecule reactant, a hairpin loop, an annealing region, a coding region, and a primer binding site. This type of protocol may be used to synthesize a wide variety of templates, in particular, H type templates useful in the practice of the invention. [0179] An efficient method to synthesize a large variety of templates is to use a "split poor' technique. The oligonucleotides are synthesized using standard 3' to 5' chemistries. First, 20 the constant 3' end is synthesized. This is then split into n different vessels, where n is the number of different codons to appear at that position in the template. For each vessel, one of the n different codons is synthesized on the (growing) 5' end of the constant 3' end. Thus, each vessel contains, from 5' to 3', a different codon attached to a constant 3' end. The n vessels are then pooled, so that a single vessel contains n different codons attached to the constant 3' end. 25 Any constant bases adjacent the 5' end of the codon are now synthesized. The pool then is split into n different vessels, where m is the number of different codons to appear at the next (more 5') position of the template. A different codon is synthesized (at the 5' end of the growing oligonucleotide) in each of the 7n vessels. The resulting oligonucleotides are pooled in a single vessel. Splitting, synthesizing, and pooling are repeated as required to synthesize all codons and 30 constant regions in the oligonucleotides.

WO 2004/016767 PCT/US2003/025984 - 59 IL TRANSFER UNITS [01801 A transfer unit.comprises an oligonucleotide containing an anti-codon sequence and a reactive unit. The anti-codons are designed to be complementary to the codons present in the template. Accordingly, the sequences used in the template and the codon lengths should be 5 considered when designing the anti-codons. Any molecule complementary to a codon used in the template may be used, including natural'or non-natural nucleotides. In certain embodiments, the codons include one or more bases found in nature (i.e., thymidine, uracil, gunidine, cytosine,,and adenine). Thus, the anti-codon can include one or more nucleotides normally found in Nature with a base, a sugar, and an optional phosphate group. Alternatively, the bases 10 may be connected via a backbone other than the sugar-phosphate backbone normally found in Nature (e.g., non-natural nucleotides). [0181] As discussed above, the anti-codon is associated with a particular type of reactive unit to form a transfer unit. The reactive unit may represent a distinct entity or may be part of the functionality of the anti-codon unit. In certain embodiments, each anti-codon sequence is 15 associated with one monomer type. For example, the anti-codon sequence ATTAG may be associated with a carbamate residue with an isobutyl side chain, and the anti-codon sequence CATAG may be associated with a carbamate residue with a phenyl side chain. This one-for-one mapping of anti-codon to monomer units allows the decoding of any polymer of the library by sequencing the nucleic acid template used in the synthesis and allows synthesis of the same 20 polymer or a related polymer by knowing the sequence of the original polymer. By changing (e.g., mutating) the sequence of the template, different monomer units may be introduced, thereby allowing the synthesis of related polymers, which can subsequently be selected and evolved. In certain preferred embodiments, several anti-codons may code for one monomer unit as is the case in Nature. 25 [01821 In certain other embodiments, where a small molecule library is to be created rather than a polymer library, the anti-codon generally is associated with a reactive unit or reactant used to modify a small molecule scaffold. In certain embodiments, the reactant is linked to the anti-codon via a linker long enough to allow the reactant to come into reactive proximity with the small molecule scaffold. The linker preferably has a length and composition to permit 30 intramolecular reactions but yet minimize intennolecular reactions. The reactants include a variety of reagents as demonstrated by the wide range of reactions that can be utilized in nucleic WO 2004/016767 PCTfUS2003/025984 -60 acid-templated synthesis (see, Examples 2,4 and 7) and can be any chemical group, catalyst (e.g., organometallic compoundss, or reactive mnoiety (e.g.., electrophiles, nucleophiles) known in the chemical. [01831 . Additionally, the association between the anti-codon and the reactive unit, for 5 example, a monomer unit or reactant, iin the transfer unit may be covalent or non-covaleht, The association maybe through a covalent bond and, in certain embodiments, the tovalent bond may be severable, 101841 Thus, the anti-codon can be associated with the reactant through a linker moiety (see Example ,). The linkage can be cleavable b' light, oxidation, hydrolysis, exposure to acid, 10 exposure to base, reduction, etc. Fruchtel et al. (1996) ANGEW. CHEM. INT. ED. ENGL. 35: 17 describes a variety of linkages useful in the practice of the invention. The linker facilitates contact of the reactant with the small molecule scaffold and in certain embodimentsidepending on the desired reaction, positions DNA as a leaving group ("autocleavable" strategy), or may link reactive groups to the template via the "scarless" linkdi strategy (which yields product 15 without leaving behind an additional atom or atoms having chemical functionality), or a "useful scar" strategy (in which a portion of the linker is left behind to be fuinctionalized in subsequent steps following linker cleavage). [0185] With the "autocleavable" linker strategy, the DNA-reactive group bond is cleaved as a natural consequence of the reaction. In the "scarless" linker strategy, DNA-templated 20 reaction of one reactive group is followed by cleavage of the linker attached through a second reactive group to yield products without leaving behind additional atoms capable of providing chemical functionality. Alternatively, a "useful scar" may be utilized on the theory that it may be advantageous to introduce useful atoms and/or chemical groups as a consequence of linker cleavage. In particular, a "useful scar" is left behind following linker cleavage and can be 25 functionalized in subsequent steps. [01861 The anti-codon and the reactive unit (monomer unit or reactant) may also be associated through non-covalent interactions such as ionic, electrostatic, hydrogen bonding, van der Waals interactions, hydrophobic interactions, pi-stacking, etc. and combinations thereof To give but one example, an anti-codon may be linked to biotin, and a monomer unit linked to 30 streptavidin. The propensity of streptavidin to bind biotin leads to the non-covalent association between the anti-codon and the monomer unit to form the transfer unit.

WO 2004/016767 PCT/US2003/025984 -61 {0187 The specific annealing of transfer units to templates permits the use of transfer uuiits at concentrations lower than concentrations used in many traditional organic syntheses. Tjhps, transfer units can be used at submillimolar concentrations (e.g. less than 100 piM, less than 10 iM, less than 1 M less than 100 nM, or less than 10 nM). 5 III. CHEMICAL REACTIONS [01881 A variety of compounds and/or libraries can be prepared using the methods described herein. In certain embodiments, compounds that are not, or do not resemble, nucleic acids or analogs thereof, are synthesized' according t5 the method of the invention In certain other embodiments, compounds that are not, or do not resemble, proteins, peptides, or analogs 10 thereof, are synthesized according to the method of the invention. (i) Coupling Reactions for Small Molecule Synthesis [0189] In some embodiments, it is possible to- create compounds such as small molecules using the methods described herein. These small molecules may be like natural products, non polymeric, and/or non-oligomeric. The substantial interest in small molecules is due in part to 15 their use as the active ingredient in many pharmaceutical preparations although they may also be used, for example, as catalysts, materials, or additives. [01901 In synthesizing small molecules using the method of the present invention, an evolvable template also is provided. The template can include a small molecule scaffold upon which the small molecule is to be built, or a small molecule scaffold may be added to the 20 template. The small molecule scaffold can be any chemical compound with two or more sites for functionalization. For example, the small molecule scaffold can include a ring system (e.g., the ABCD steroid ring system found in cholesterol) with functionalizable groups coupled to the atoms making up the rings. In another example, the small molecule may be the underlying structure of a pharmaceutical agent such as morphine, epothilone or a cephalosporin antibiotic. 25 The sites or groups to be functionalized on the small molecule scaffold may be protected using methods and protecting groups known in the art. The protecting groups used in a small molecule scaffold may be orthogonal to one another so that protecting groups can be removed one at a time. [0191] In this embodiment, the transfer units comprise an anti-codon associated with a 30 reactant or a building block for use in modifying, adding to, or taking away from the small WO 2004/016767 PCT/US2003/025984 - 62 molecule scaffold The reactants or building block may be, for example, electrophiles (e.g., acetyl, amides, acid chlorides, ester§, nitriles, imines), nhcleophiles (e.g., amines, hydroxyl groups, thiols) catalysts (e.g., organotetallic catalysts), or side chains. Tlie transfer units are allowed to contact the, template under hydridizidg conditions. As a result of oligonucleotide 5 annealing,'the attached reactant or building block is allowed to react with a site on the small molecule scaffold. In certain embodiments, protecting groups on the small nwolecule teinplate are removed one at a time from the sites to- be fdnctionaiized so that the reactant of the transfer unit will 'eac-t at only the desired position on the scaffold. 101921 The reaction conditions, linker, reactant, and site to be fiuctionalized are chosen 10 to avoid intermolecular reactions and accelerate intramolecular reactions. Sequential or simultaneous contacting of the template with transfer units can be employed depending on the particular compound to be synthesized. In certain embodiments of special interest, the multi-step synthesis of chemical compounds is provided in which the template is contacted sequentially with two or more transfer units to facilitate rniulti-step synthesis of complex chemical. 15 compounds. [01931 After the sites on the scaffold have-been modified, the newly synthesized small molecule remains associated with the template that encoded its synthesis. Decoding the sequence of the template permits the deconvolutibn of the synthetic history and thereby' the structure of the small molecule. The template can also be amplified in order to create more of 20 the desired small molecule and/or the template can be evolved (mutagenized) to create related small molecules. The small molecule can also be cleaved from the template for purification or screening. (i) Coupling Reactions for Polymer Synthesis [0194] In certain embodiments, polymers, specifically unnatural polymers, are prepared 25 according to the method of the present invention. The unnatural polymers that can be created using the inventive method and system include any unnatural polymers. Exemplary unnatural polymers include, but are not limited to, peptide nucleic acid (PNA) polymers, polycarbamates, polyureas, polyesters, polyacrylate, polyalkylene (e.g., polyethylene, polypropylene), polycarbonates, polypeptides with unnatural stereochemistry, polypeptides with unnatural amino 30 acids, and combination thereof. In certain embodiments, the polymers comprise at least 10, 25, WO 2004/016767 PCT/US2003/025984 -63 75, 100, 125, 150 monomer units or more. The polymers synthesized using the inventive system may be used, for example, as catalysts, pharmaceuticals, metal chelators, or catalysts. [0195]q In preparing certain unnatural polymers, the monomer units attached to the anti codons may be any monomers or oligomers capable of being joined together to form a polymer. 5 The mononer units may be, for example,.carbarnates, D-amino acids, unnatural amino acids, PNAs, ureas, hydroxy acids, esters, carbonites, acrylates; or ethers. In certain embodiments, the monomer units have two reactive groups used to link the monomer unit into the'growing polymer chain, as depicted in Figure 4, Preferably, the two reactive groups arenot the same so that the monomer unit may be incorporated into the polymer in a directional sense, for example, 10 at one end may be an electrophile and at the other end a nucleophile. Reactive groups may include, but are not limited to, esters, amides, carboxylic acids, activated carbonyl groups, acid chlorides, amines, hydroxyl groups, and thiols. In certain embodiments, the reactive groups are masked or protected (Greene et al. (1999) PROTECTIVE GRouPs N ORGANIC SYNTHESIS 3rd Edition; Wiley) so that polymerization may not take place until a desired time when the reactive 15 groups are deprotected. Once the monomer units are assembled along the nucleic acid template, initiation of the polymerization sequence results in a cascade of polymerization and deprotection steps wherein the polymerization step results in deprotection of a reactive group to be used in the subsequent polymerization step. (0196] The monomer units to be polymerized can include two or more monomers 20 depending on the geometry along the nucleic acid template. The monomer units to be polymerized must be able to stretch along the nucleic acid template and particularly across the distance spanned by its encoding anti-codon and optional spacer sequence. In certain embodiments, the monomer unit actually comprises two monomers, for example, a dicarbamate, a diurea, or a dipeptide. In yet other embodiments, the monomer unit comprises three or more 25 monomers. Example 9C, for example, discloses the synthesis of PNA based polymers wherein each monomer unit comprises four PNA molecules. [0197] The monomer units may contain any chemical groups known in the art. Reactive chemical groups especially those that would interfere with polymerization, hybridization, etc., are preferably masked using known protecting groups (Greene et al. (1999) supra). In general, 30 the protecting groups used to mask these reactive groups are orthogonal to those used in protecting the groups used in the polymerization steps.

WO 2004/016767 PCT/US2003/025984 -64' 101981 It has been discovered that, under 'ertain -circumstances, the type of chemical reaction inay affect the fidelity of fhe polymerization process. For example, distance independent chemical reactions, (for example, reactions that occur efficietly when the reactive, units hre spaced apart by intervening bases, for'example, amine acylatibn reactions) may result in S the spuridus incorporation of the wrong monomers at a particular position of a polymer chain. In contrast, by choosing chemical reactions for template mediated syntheses tlzat are distance dependent (for example, reactions that becoii inefficient the further the reactive units are spaced part via intervening bases, fot example, reduqctive amination reactions), it is possible control the fidelity of the polymerization process. Example 9 discusses in detail effect of using; 10 distance dependent chemical reactions to enhance the fidelity of the polymerization process during template mediated synthesis. (iii) Functional Group Transformations 101991 Nucleic acid-templated synthesis can be used to effect functional group transformations that either (i) unmask or (ii) interconvert"functionality used in coupling 15 reactions. By- exposing or creating a reactive group within a sequence-programmed subset of a library, nucleic acid-templated functional group interconversions permit the generation of library diversity by sequential unmasking. The sequential unmasking approach offers the major advantage of enabling reactants that would nornially lack the ability to be linked to a niucleic acid (for example, simple alkyl halides) to contribute to library diversity by reacting with a sequence 20 specified subset of templates in an intermolecular, non-templated reaction mode. This advantage significantly increases the types of structures that can be generated. [02001 One embodiment of the invention involves deprotection or unmasking of functional groups present in a reactive unit. According to this embodiment, a nucleic acid template is associated with a reactive unit that contains a protected functional group. A transfer 25 unit, comprising an oligonucleotide complimentary to the template codon region and a reagent capable of removing the protecting group, is annealed to the template, and the reagent reacts with the protecting group, removing it from the reactive unit. To further functionalize the reactive unit the exposed functional group then is subjected to a reagent not linked to a nucleic acid. In some embodiments, the reactive unit contains two or more protected functional groups. In still 30 other embodiments, the protecting groups are orthogonal protecting groups that are sequentially removed by iterated annealing with reagents linked to transfer units.

WO 2004/016767 PCT/US2003/025984 -65 [0201] Another embodiment of the invention involves interconversions of functional. groups present on a reactive unit. According to ihis embodiment, a transfer unit associatedwith a reagent that c'an catalyze a reaction'is annealed to'a' template bearing thd'reactive unit. A reagent not linked to a nucleic acid is added to the. reaction, and the transfer unit reagent 5 catalyzes the reaction between the unlinked reagent and the reactive unit, yielding a nevly functionalized reactive unit. In some' embodiments, the reactive unit contains two or niore functional groups which are sequentially intetc onveried'by iterative exposure to different transfer unit-botufd teagents. (iv) Reaction Conditions 10 [0202] Nucleic acid-templated reactions can occur in aqueous or non-aqueous (i.e., organic) solutions, or a mixture of one or more aqueous and non-aqueous solutions. In aqueous solutions, reactions can be performed at pH ranges from about 2 to abbut 12, or preferably from about 2 to about 10, or.more preferably from about 4 to about 10. The reactions used in DNA templated chemistry preferably should not require very basic conditions (e.g., pH > 12,,pH > 10) 15 or very acidic conditions (e.g., pH < 1, pH <2, pH < 4), because extreme conditions may'lead to degradation or modification of the nucleic acid template and/or molecule (for example, the polymer, or small molecule) being synthesized. The aqueous solution can contain one or more inorganic salts, including, but not limited to, NaC1, Na 2 S04, KCI, Mg 4 2 , Mn 2 , etc., at various concentrations. 20 [02031 Organic solvents suitable for nucleic acid-templated reactions include, but are not limited to, methylene chloride, chloroform, dimethylfonamide, and organic alcohols, including methanol and ethanol. To permit quantitative dissolution of reaction components in organic solvents, quaternized ammonium salts, such as, for example, long chain tetraalkylammonium salts, can be added (Jost et al. (1989) NUCLEIC ACIDs RES. 17: 2143; MePnikov et al. (1999) 25 LANGMUTR 15: 1923-1928), [0204] Nucleic acid-templated reactions may require a catalyst, such as, for example, homogeneous, heterogeneous, phase transfer, and asymmetric catalysis. In other embodiments, a catalyst is not required. The presence of additional, accessory reagents not linked to a nucleic acid are preferred in some embodiments. Useful accessory reagents can include, for example, 30 oxidizing agents (e.g., NatO 4 ); reducing agents (e.g., NaCNBH 3 ); activating reagents (e.g., EDC, NHS, and sulfo-NHS); transition metals such as nickel (e.g., Ni(NO3) 2 ), rhodium (e.g. RhCl 3

),

WO 2004/016767 PCT/US2003/025984 -66 ruthenium (e.g. RuC 3 ), copper (e.g. Cu(NO 3

)

2 ), cobalt (e.g. CoCl 2 ), iron (e.g. Fe.(NO 3

)

3 ), osmium (e.g. Os04), titanium (e.g. TiC4 or titanium tetraisopropoxide), palladium (e.g. NaPdC 4 ), or Ln; transition metal ligands (e.g., phosphines, amines, and halides); Lewis' acids; and Lewis bases. ,. 5 [0205] Reaction conditions preferably are optimized to suit the nature of the reactive units and oligonucleotides used. (v) Classes of Chemical Reactions [0206] Known chemical reactions for synthesizing polyiers, small molecules, or other chemical compounds can be used in nucleic acid-templated reactions. Thus, reactions such as 0 those listed in March's Advanced Organic Chemistry, Organic Reactions, Organic Syntheses, organic text books, journals such as Journal of the American Chemical Society, Journal of Organic Chemistry, Tetrahedron, etc., and Carruther's Some Modern Methods of Organic Chemistry can be used. The chosen reactions preferably are compatible with nucleic acids such as DNA or RNA or are compatible with the modified nucleic acids used as the template. 5 [0207] Reactions useful in nucleic-acid templated chemistry include, for example, substitution reactions, carbon-carbon bond forming reactions, elimination reactions, acylation reactions, and addition reactions. An illustrative but not exhaustive list of aliphatic nucleophilic substitution reactions useful in the present invention includes, for example, SN2 reactions, SN1 reactions, SNi reactions, allylic rearrangements, nucleophilic substitution at an aliphatic trigonal 0 carbon, and nucleophilic substation at a vinylic carbon. [0208] Specific aliphatic nucleophilic substitution reactions with oxygen nucleophiles include, for example, hydrolysis of alkyl halides, hydrolysis of gen-dihalides, hydrolysis of 1,1,1-trihalides, hydrolysis of alkyl esters or inorganic acids, hydrolysis of diazo ketones, hydrolysis of acetal and enol ethers, hydrolysis of epoxides, hydrolysis of acyl halides, 25 hydrolysis of anhydrides, hydrolysis of carboxylic esters, hydrolysis of amides, alkylation with alkyl halides (Williamson Reaction), epoxide formation, alkylation with inorganic esters, alkylation with diazo compounds, dehydration of alcohols, transetherification, alcoholysis of epoxides, alkylation with oniurn salts, hydroxylation of silanes, alcoholysis of acyl halides, alcoholysis of anhydrides, esterfication of carboxylic acids, alcoholysis of carboxylic esters 30 (transesterfication), alcoholysis of aides, alkylation of carboxylic acid salts, cleavage of ether with acetic anhydride, alkylation of carboxylic acids with diazo compounds, acylation of WO 2004/016767 PCT/US2003/025984 -. ' -67 caroxylic acids. with acyl halides; acylatioi of arioxylic acids with carboxylic acids, formation, of oxoninim salts, preparation of peroxides'and hydrop6roxides, preparation of inbrganit esters (e.g., nitrites, nitrates, -sulfonates), prparation of alcohpls from amines, and preparation of mixed organic-inorganic anhydrides. 5 [02091 Specific aliphatic nueleophilic substitution reactions with sulfudr nucleophiles, which tend to be better nucleophiles than their oxygen- analogs, include, for'example, attack by SH at an alkyl carbon to form thiols, attack byS at an alkyl carbon to form thioethers, attack by SH or SR at an acyl carbon, formation of disulfide§, formation of Bunte salts, alkylatiori of, sulfinic acid salts, and formation of alkyl thiocyanates. 10 [02101 Aliphatic nucleophilic substitution reactions with nitrogen nucleophiles include, for example, alkylation of amines, N-arylation'of amines,. replacement of a hydroxy by an amino group, trahsamination, tiansamidation, alkylati6n of mines with diazo compounds, amination ,of epoxides, amination of oxetanes, amination of aziridines, amination of alkanes, formation of isocyanides, acylation of amines by acyl halides, acylation of amines by anhydrides, aqylation of 15 amines by carboxylic acids, acylation of amines by carboxylic esters, acylation of amines by aides, acylation of amines by other acid derivatives, N-alkylation or N-arylation'of amides and imides, N-acylation of amides and imides, formation of aziridines from epoxides, formation of nitro compounds, formation of azides, formation of isocyanates and isothiocyanates, and formation of azoxy compounds. 20 102111 Aliphatic nucleophilic substitution reactions with halogen nucleophiles include, for example, attack at an alkyl carbon, halide exchange, formation of alkyl halides from esters of sulfuric and sulfonic acids, formation of alkyl halides from alcohols, formation of alkyl halides from ethers, formation of halohydrins from epoxides, cleavage of carboxylic esters with lithium iodide, conversion of diazo ketones to a-halo ketones, conversion of amines to halides, 25 conversion of tertiary amines to cyanamides (the von Braun reaction), formation of acyl halides from carboxylic acids, and formation of acyl halides from acid derivatives. [0212] Aliphatic nucleophilic substitution reactions using hydrogen as a nucleophile include, for example, reduction of alkyl halides, reduction of tosylates, other sulfonates, and similar compounds, hydrogenolysis of alcohols, hydrogenolysis of esters (Barton-McCombie 30 reaction), hydrogenolysis of nitriles, replacement of alkoxyl by hydrogen, reduction of epoxides, reductive cleavage of carboxylic esters, reduction of a C-N bond, desulfurization, reduction of VVU 404/1016767 'PCT/US2003/025984 -68 acyl halides, reduction of carboxylic acids, esters, and anhydrides to aldehydes, and reduction of ainides to aldehydes. [02131, Although certain carbon nucleophiles may be too nucleophilic and/or basic to be used in certain embodiments of the invention, aliphatic nhtleophilic substitution reactions usiiig 5 carbon nucleophiles include, for example, coupling with silanes, coupling of alkyl halides (the Wurtz reaction), the reaction of alkyl halides and sulfonate esters with Group I (I A) and II (1 A) organometallic reagents, reaction of alkyl halides and sulfonate esters with organocuprates, reaction of akyl halides and sulfonate esters with other organometallic reagents, allylic and propargylic coupling with a halide substrate, coupling of organometallic reagents with esters of 10 sulfuric and sulfonic acids, sulfoxides, and sulfones, coupling involving alcohols, coupling of organometallic reagents with carboxylic esters, coupling of organometallic reagents with compounds containing an esther linkage, reaction of organometallic reagents with epoxides, reaction of organometallics with aziridide, alkylation at a carbon bearing an active hydrogen, alkylation of ketones, nitriles, and carboxylic esters, alkylation of carboxylic acid salts, .5 alkylation at a position a to a heteroatom (alkylation of 1,3-dithianes), alkylation of dihydro-1,3 oxazine (the Meyers synthesis of aldehydes, ketones, and carboxylic acids), alkylation with trialkylboranes, alkylation at an alkynyl carbon, preparation of nitriles, direct conversion of alkyl halides to aldehydes and ketones, conversion of alkyl halides, alcohols, or alkanes to carboxylic acids and their derivatives, the conversion of acyl halides to ketones with organometallic 0 compounds, the conversion of anhydrides, carboxylic esters, or aides to ketones with organometallic compounds, the coupling of acyl halides, acylation at a carbon bearing an active hydrogen, acylation of carboxylic esters by carboxylic esters (the Claisen and Dieckmann condensation), acylation of ketones and nitriles with carboxylic esters, acylation of carboxylic acid salts, preparation of acyl cyanides, and preparation of diazo ketones, ketonic 5 decarboxylation. [0214] Reactions which involve nucleophilic attack at a sulfonyl sulfur atom may also be used in the present invention and include, for example, hydrolysis of sulfonic acid derivatives (attack by OH), formation of sulfonic esters (attack by OR), formation of sulfonamides (attack by nitrogen), formation of sulfonyl halides (attack by halides), reduction of sulfonyl chlorides 0 (attack by hydrogen), and preparation of sulfones (attack by carbon).

WO 20041016767 PCTfUS2003/025984 102151 Momatic electrophilic substitution'reactions may also be used in nucleotide templated chemistry.; Hydrogen exchalgereactions ar6 examples of aromatic electropilic substitution reactions that use hydrogen as the electrophile. Aromatic e6trophilic substitution reactions which use nitrogen electrophiles include, for example, nitration and nitro-de 5 hydrogenation, nitr6sation of nitroso-de-hydrogenation, diazonium coupling,' direct introduction of the diazonidm group, and amination-or ariino-de--hydrogenation. Reactions of this type with sulfur electrophiles include, for example, sulfohation, suilfo-de-hydrogenation, halosulfonation, halosulfo-de-hydrogenation, sulfurization, and sulfonylation. Reactions using halogen electrophiles include, for example, halogenation, and halo-de-hydrogenation. Aromatic 10 electrophilic 'substitution reactions with carbon electrophiles include, for example, Friedel-Crafts alkylation, alkylation, alkyl-de-hydrogenation, Friedel-Crafts arylation (the Scholl reaction), Friedel-Crafts acylation, formylation with disubstituted formamides, formylation with zinc cyanide and HCI (the Gatterman reaction), formylation with chloroform (the Reimer-Tiemann reaction), other formylations, formyl-de-hydrogenation, carboxylation with carbonyl halides, 15 carboxylation with carbon dioxide (the Kolbe-Schmitt reaction), amidation with isocydnates, N alkylcarbamoyl-de-hydrogenation, hydroxyalkylation, hydroxyalkyl-de-hydrogenation, cyclodehydration of aldehydes and ketones, haloalkylation, halo-de-hydrogenation, aminoalkylation, amidoalkylation, dialkylaminoalkylation, dialkylamino-de-hydrogenation, thioalkylation, acylation with nitriles (the Hoesch reaction), cyanation, and cyano-de 20 hydrogenation. Reactions using oxygen electrophiles include, for example, hydroxylation and hydroxy-de-hydrogenation. [0216] Rearrangement reactions include, for example, the Fries rearrangement, migration of a nitro group, migration of a nitroso group (the Fischer-Hepp Rearrangement), migration of an arylazo group, migration of a halogen (the Orton rearrangement), migration of an alkyl group, 25 etc. Other reaction on an aromatic ring include the reversal of a Friedel-Crafts alkylation, decarboxylation of aromatic aldehydes, decarboxylation of aromatic acids, the Jacobsen reaction, deoxygenation, desulfonation, hydro-de-sulfonation, dehalogenation, hydro-de-halogenation, and hydrolysis of organometallic compounds. [0217] Aliphatic electrophilic substitution reactions are also useful. Reactions using the 30 SEL, SE2 (front), SE 2 (back), Sii, addition-elimination, and cyclic mechanisms can be used in the present invention. Reactions of this type with hydrogen as the leaving group include, for WO 2004/016767 PCT/US2003/0259l4 -70 example, hydrogen exchange (deuterio-de-hydrogenatiol, deuteriation), migration.of a double bond, and keto-enol tautomerization. Reactions with halogen electrophiles include, for example, halogenation of aldehydes and ketones, halogenation of caboxylic acids and acyl halides, and halogenation of sulfoxides and sulfones. Reactions with nitrogen electrophiles include, for 5 example, aliphatic diazonium coupling, nitrosation at a carbon bearing an active hydrogen, direct formation of diazo compounds, conversion of amides to a-azido amides, direct arnination at an activated position, and insertion by nitrenes. Reactions with sulfur or selenium electrophiles include, for example, sulfenylation, sulfonation, and1 selenylation of ketones and carboxylic esters. Reactions with carbon electrophiles include, for example, acylation at an aliphatic 10 carbon, conversion of aldehydes to p-keto esters or ketones, cyanation, cyano-d&-hydrogenlation, alkylation of alkanes, the Stork enamine reaction, and insertion by carbenes. Reactions with metal electrophiles include, for example, metalation with organometallic compounds, metalation with metals and strong bases, and conversion of enolates to silyl enol ethers. Aliphatic electrophilic substitution reactions with metals as leaving groups include, for example, 15 replacement of metals by hydrogen, reactions between organometallic reagents and oxygen, reactions between organometallic reagents and peroxides, oxidation of trialkylboranes to borates, conversion of Grignard reagents to sulfur compounds, halo-de-metalation, the conversion of organometallic compounds to amines, the conversion of organometallic compounds to ketones, aldehydes, carboxylic esters and amides, cyano-de-metalation, transmetalation with a metal, 20 transmetalation with a metal halide, transmetalation with an organometallic compound, reduction of alkyl halides, metallo-de-halogenation, replacement of a halogen by a metal from an organometallic compound, decarboxylation of aliphatic acids, cleavage of alkoxides, replacement of a carboxyl group by an acyl group, basic cleavage of p-keto esters and $ diketones, haloform reaction, cleavage of non-enolizable ketones, the Haller-Bauer reaction, 25 cleavage of alkanes, decyanation, and hydro-de-cyanation. Electrophlic substitution reactions at nitrogen include, for example, diazotization, conversion of hydrazines to azides, N-nitrosation, N-nitroso-de-hydrogenation, conversion of amines to azo compounds, N-halogenation, N-halo de-hydrogenation, reactions of amines with carbon monoxide, and reactions of amines with carbon dioxide. 30 [02181 Aromatic nucleophilic substitution reactions may also be used in the present invention. Reactions proceeding via the SNAr mechanism, the SN 1 mechanism, the benzyne mechanism, the SRN1 mechanism, or other mechanism, for example, can be used. Aromatic WO 2004/016767 PCT/US2003/025984 -71 nucleophilic substitution reactions with. oxygen nucleophiles. include, for example,. hydroxy-de-. halogenation, alkali fusion of sulfonate salts, and replacement of OR or OAr. Reactions with s4lfui nuclepphiles include, for example, replacement by'SH or SR. Reactions using nitrogen nucleophiles include, for example, replacement by NH 2 , NHR, or NR 2 , and replacement of a 5 hydroxy group by an amino group: Reactions with halogen nucleophiles include, for example, the introduction halogens. Aromatic nucleophilic substitution reactions with hydrogen as the nucleophile include, for example, reduction of phenols and phenolic esters and ethers, and reduction of halides and nitro compounds. Reactions with carbon nucleophiles iriclude, for example, the Rosenmund-von Braun reaction, coupling of orgaiometallic compounds with aryl 10 halides, ethers, and carboxylic esters, arylation at a carbon containing an active hydrogen, conversions of aryl substrates to carboxylic acids, their derivatives, aldehydes, dnd ketones, and the Ullmann reaction. Reactions with hydrogen as the leaving group include, for example, alkylation, arylation, and amination of nitrogen heterocycles. Reactions with N2+ as the leaving group include, for example, hydroxy-de-diazoniation, replacement by sulfur-containing groups, 15 iodo-de-diazoniation, and the Schiemann reaction. Rearrangement reactions include, for example, the von Richter rearrangement, the Sommelet-Hauser rearrangement, rearrangement of aryl hydroxylamines, and the Smiles rearrangement. [0219] Reactions involving free radicals can also be used, although the free radical reactions used in nucleotide-templated chemistry should be carefully chosen to avoid 20 modification or cleavage of the nucleotide template. With that limitation, free radical substitution reactions can be used in the present invention. Particular free radical substitution reactions include, for example, substitution by halogen, halogenation at an alkyl carbon, allylic halogenation, benzylic halogenation, halogenation of aldehydes, hydroxylation at an aliphatic carbon, hydroxylation at an aromatic carbon, oxidation of aldehydes to carboxylic acids, 25 formation of cyclic ethers, formation of hydroperoxides, formation of peroxides, acyloxylation, acyloxy-de-hydrogenation, chlorosulfonation, nitration of alkanes, direct conversion of aldehydes to aides, amidation and amination at an alkyl carbon, simple coupling at a susceptible position, coupling of alkynes, arylation of aromatic compounds by diazonium salts, arylation of activated alkenes by diazonium salts (the Meerwein arylation), arylation and 30 alkylation of alkenes by organopalladium compounds (the Heck reaction), arylation and alkylation of alkenes by vinyltin compounds (the Stille reaction), alkylation and arylation of aromatic compounds by peroxides, photochemical arylation of aromatic compounds, alkylation, WO 2004/016767 PCT/US2003/025984 -72 acylation, and carbalkoxylation of nitrogen heteroccles Particular reactions in which NIF is the leaving group include, for example, replacement.bf the diazonium group by hydrogen, replacement of the diazonium group by chlorine or bromine, nitro-de-diazbniation, replacement of the diazonium group-by sulfur-containing groups, aryldimerization with diazonium salts, 5 methylation of diazonium salts, vinylation of diazonium salts, arylation of diazonium salts, and conversion of diazonium salts to aldehydes, kdtones, or carboxylic acids. Free radical substitution reactions with metals as leavirig groups iclUide, for example; coupling of Grignard reagents,'coupling of boranes, and coupling of other organometallic reagents: Reaction with halogen as the leaving group are included. Other, free radical substitution reactions with various 10 leaving groups include, for example, desulfurization with Raney Nickel, conversion of sulfides to organolithium compounds, decarboxylative dimerization (the Kolbe reaction), the Hunsdiecker reaction, decarboxylative allylation, and decarbonylation of aldehydes and acyl halides. [0220] Reactions involving additions to carbon-carbon multiple bonds are also used in 15 nucleotide-templated chemistry. Any mechanism may be used in the addition reaction including, for example, electrophilic addition, nucleophilic addition, free radical addition, and cyclic mechanisms. Reactions involving additions to conjugated systems can also be used. Addition to cyclopropane rings can also be utilized. Particular reactions include, for example, isomerization, addition of hydrogen halides, hydration of double bonds, hydration of triple bonds, addition of 20 alcohols, addition of carboxylic acids, addition of H 2 S and thiols, addition of arnmonia and amines, addition of amides, addition of hydrazoic acid, hydrogenation of double and triple bonds, other reduction of double and triple bonds, reduction of the double and triple bonds of conjugated systems, hydrogenation of aromatic rings, reductive cleavage of cyclopropanes, hydroboration, other hydrometalations, addition of alkanes, addition of alkenes and/or alkynes to 25 alkenes and/or alkynes (e.g., pi-cation cyclization reactions, hydro-alkenyl-addition), ene reactions, the Michael reaction, addition of organometallics to doable and triple bonds not conjugated to carbonyls, the addition of two alkyl groups to an alkyne, 1,4-addition of organometallic compounds to activated double bonds, addition of boranes to activated double bonds, addition of tin and mercury hydrides to activated double bonds, acylation of activated 30 double bonds and of triple bonds, addition of alcohols, anines, carboxylic esters, aldehydes, etc., carbonylation of double and triple bonds, hydrocarboxylation, hydroformylation, addition of aldehydes, addition of HCN, addition of silanes, radical addition, radical cyclization, WO 2004/016767 PCT/U1S2003/025984 -73 halogbnation of double and triple bonds (addition of halogen,.halogen), halolactonization, halolactaniization, addition of hypohalous acids and hypohalites (addition of halogen, oxygen), addition of sulfur compounds (addition of halogen, sulfur), addition of halogen and an amino group (addition of halogen, nitrogen), addition of NOX and NO 2 X (addition of halogen, . 5 nitrogen), addition of XN 3 (addition of halogen, nitrogen), addition of alkyl halides (addition of halogen, carbon), addition of acyl halides (addition of halogen, carbon), hydroxylation (addition of oxygen, oxygen) (e.g., asymmetric dihydroxylation reaction with OsO4), dihydroxylation of aromatic rings, epoxidation (addition of oxygen, oxygen) (e.g., Sharpless asymmetric epoxidation), photooxidation of dienes (addition of oxygen, oxygen), hydroxysulfenylation 10 (addition of oxygen, sulfur), oxyamination (additionlof oxygen, nitrogen), diamiiation (addition of nitrogen, nitrogen), formation of aziridines (addition of nitrogen), aminosulfeAylation (addition of nitrogen, sulfur), acylacyloxylation and acylamidation (addition of oxygen, carbon or nitrogen, carbon), 1,3-dipolar addition(addition of oxygen, nitrogen, carbon), Diels-Alder reaction, heteroatom Diels-Alder reaction, all carbon .3 +2 cycloadditions, dimerization of 15 alkenes, the addition of carbenes and carbenoids to double and triple bonds, trimerization and tetramerization of alkynes, and other cycloaddition reactions. [02211 , In addition to reactions involving additions to carbon-carbon multiple bonds, addition reactions to carbon-hetero multiple bonds can be used in nucleotide-templated chemistry. Exemplary reactions include, for example, the addition of water to aldehydes and 20 ketones (formation of hydrates), hydrolysis of carbon-nitrogen double bond, hydrolysis of aliphatic nitro compounds, hydrolysis of nitriles, addition of alcohols and thiols to aldehydes and ketones, reductive alkylation of alcohols, addition of alcohols to isocyanates, alcoholysis of nitriles, formation of xanthates, addition of H1 2 S and thiols to carbonyl compounds, formation of bisulfite addition products, addition of amines to aldehydes and ketones, addition of aides to 25 aldehydes, reductive alkylation of ammonia or amines, the Mannich reaction, the addition of amines to isocyanates, addition of ammonia or amines to nitriles, addition of amines to carbon disulfide and carbon dioxide, addition of hydrazine derivative to carbonyl compounds, formation of oximes, conversion of aldehydes to nitriles, formation of gem-dihalides from aldehydes and ketones, reduction of aldehydes and ketones to alcohols, reduction of the carbon-nitrogen double 30 bond, reduction of nitriles to amines, reduction of nitrites to aldehydes, addition of Grignard reagents and organolithium reagents to aldehydes and ketones, addition of other organometallics to aldehydes and ketones, addition of trialkylallylsilanes to aldehydes and ketones, addition of WO 2004/016767 PCT/US2003/025984 -74 conjugated alkenes to aldehydes (the Baylis-Hillnan reaction), the Reformatsky reaction, the conversion of carboxylic acid salts to kdtones with organometallic compounds, the addition of Grignard reagents to"acid derivatives, the addition of organometallic compdunds to CO 2 and CS 2 ,, addition of organometallic compounds to C=N compounds, addition of 6arbenes and 5 diazoalkans to C=N compounds, addition of Grignard reagents to nitriles and'isocyanales, the Aldol reaction, Mukaiyama Aldol and'related reactions, Aldol-type reactions, between carboxylic esters or amides and aldehydes or ketones, the 4noeven 1l reaction (e.g., the Nef reaction, the Favorskii'reaction), the Peterson alkeiylation reaction, the addition of active hydrogen compounds to CO 2 and CS 2 , the Perkin reaction; Darzens glycidic ester condensation, the 3 Tollens' reaction, the Wittig reaction, the Tebbe'alkenylation, the Petasis alkenylation, alternative alkenylations, the Thorpe reaction, the Thorpe-Ziegler reaction, addition of silanes, formation of cyanohydrins, addition of HCN'to'C=N and CN bonds, the Prins reaction, the benzoin condensation, addition of radicals to 6=o, C=S, C=N compounds, the Ritter reaction, acylation of aldehydes and ketones, addition of aldehydes to aldehydes, the addition' of 5 isocyanates to isocyanates (formation of carbodiimides), the conversion of carboxylic acid salts to nitriles, the formation of epoxides from aldehydes and ketones, the formation of episulfides and episulfones, the formation of p-lactones and oxetanes (e.g., the Paterno-Buchi reaction), the formation of p-lactams, etc Reactions involving addition to isocyanides include the addition of water to isocyanides, the Passerini reaction, the Ug reaction, and the formation of metalated aldimines. [02221 Elimination reactions, including a, P, and y eliminations, as well as extrusion reactions, can be performed using nucleotide-templated chemistry, although the strength of the reagents and conditions employed should be considered. Preferred elimination reactions include reactions that go by El, E2, ElcB, or E2C mechanisms. Exemplary reactions include, for 5 example, reactions in which hydrogen is removed from one side (e.g., dehydration of alcohols, cleavage of ethers to alkenes, the Chugaev reaction, ester decomposition, cleavage of quarternary ammonium hydroxides, cleavage of quaternary ammonium salts with strong bases, cleavage of amine oxides, pyrolysis of keto-ylids, decomposition of toluene-p-solfonylhydrazones, cleavage of sulfoxides, cleavage of selenoxides, cleavage of sulfornes, dehydrogalogenation of alkyl 0 halides, dehydrohalogenation of acyl halides, dehydrohalogenation of sulfonyl halides, elimination of boranes, conversion of alkenes to alkynes, decarbonylation of acyl halides), reactions in which neither leaving atom is hydrogen (e.g., deoxygenation of vicinal diols, WO 2004/016767 PCT/US2003/025984 -75 cleavage of cyclic thionocarbonates, conversion of epoxides to episulfides and alkenes, the Rimberg-Backlund reaction, conversion bf aziridines to alkenes, dehalogenation of vicinal dihalides, dehalogenation of a-halo acyl halides, and elim'intion of a halogen and a hetero group), fragmentation reactions (i.e., reactions in which carbon is the positive leaving group or 5 the electrofuge, such as, for example, fragmentation of 7-amino and y-hydroxy halides, fragmentation of 1,3-diols, decarboxylation of p-hydroxy carboxylic acids, decarboxylation offp Jactones, fragmentation of ap -epoxy hydrazones, elimination of CO from briged bicyclic compounds, and elimination of CO 2 from bridged bicyclic compounds), reactions in which C=N or C=N bonds are formed (e.g., dehydration of aldoximes or similar compounds, conversion of 10 ketoximes to nitriles, dehydration of unsubstituted amides, and conversion of N-alkylformamides to isocyanides), reactions in which C=O bonds are formed (e.g., pyrolysis of P-hydroxy alkenes), and reactions in which N=N bonds are formed (e.g., eliminations to give diazoalkenes). Extrusion reactions include, for example, extrusion of N 2 from pyrazolines, extrusion of N 2 from pyrazoles, extrusion of N 2 from triazolines, extrusion of CO, extrusion of C0 2 , extrusion of SO 2 , 15 the Story synthesis, and alkene synthesis by twofold extrusion. [0223] Rearrangements, including, for example, nucleophilic rearrangements, electrophilic rearrangements, prototropic rearrangements, and free-radical rearrangements, can also be performed using nucleotide-templated chemistry. Both 1,2 rearrangements and nori-1,2 rearrangements can be performed. Exemplary reactions include, for example, carbon-to-carbon 20 migrations of R, H, and Ar (e.g., Wagner-Meerwein and related reactions, the Pinacol rearrangement, ring expansion reactions, ring contraction reactions, acid-catalyzed rearrangements of aldehydes and ketones, the dienone-phenol rearrangement, the Favorskii rearrangement, the Arndt-Eistert synthesis, homologation of aldehydes, and homologation of ketones), carbon-to-carbon migrations of other groups (e.g., migrations of halogen, hydroxyl, 25 amino, etc.; migration of boron; and the Neber rearrangement), carbon-to-nitrogen migrations of R and Ar (e.g., the Hofmann rearrangement, the Curtius rearrangement, the Lossen rearrangement, the Schmidt reaction, the Beckman rearrangement, the Stieglits rearrangement, and related rearrangements), carbon-to-oxygen migrations of R and Ar (e.g., the Baeyer-Villiger rearrangement and rearrangment of hydroperoxides), nitrogen-to-carbon, oxygen-to-carbon, and 30 sulfur-to-carbon migration (e.g., the Stevens rearrangement, and the Wittig rearrangement), boron-to-carbon migrations (e.g., conversion of boranes to alcohols (primary or otherwise), WO 2004/016767 PCT/US2003/025984 76 conversion of boranes to aldehydes, conversion 6f koraries to carboxylic acids, conversion of vinylic boranes to alkenes, formation of a kynes from bones and acetylidesformation of alkenes from boranes and acetylides, and formation of ketdnes from boranes and acetylides), electro 6 yclic rearrangements (e.g., of cyclobutenes and .1, 3-cyclohexadienes, or conversion of 5 stilbenes to' phenanthrenes),' sigmatropic rearraiigements (e.g., (1,j) sigmatropic migrations of hydrogen, (1,j) gigmatropic iigrations of carbon, conversion of vinylcyclopropanes to cyclopentenels, the Cope rearrangement, the Claisen rearr'angement, the Fischer indole synthesis, (2,3) sigmatropic rearrangements, and the benzidine rearrangement), other cyclic rearrangements (e.g., metathesis of alkenes, the di-n-methane an'drelated rean-angements, and the Hofmann 10 L5ffler and related reactions), and non-cyclic re'arrangements (e.g., hydride shifts, the Chapinan rearrangement, the Wallanh rearrangement,'and dyotropic rearrangements). [02241 Oxidative and reductive reaciops may also be performed using nucleptide templated chemistry. Exemplary reactions may involve, for example, direct electron transfer, hydride transfer, hydrogen-atom transfer, formation of ester intermediates, displacement 15 mechanisms, or. addition-elimination mechanisms. Exemplary oxidations include, for example, eliminations of hydrogen (e.g., aromatization of six-membered rings, dehydrogenations yielding carbon-carbon double bonds, oxidation or dehydrogenation of alcohols to aldehydes and ketones, oxidation of phenols and aromatic amines to quinones, oxidative cleavage of ketones, oxidative cleavage of aldehydes, oxidative cleavage of alcohols, ozonolysis, oxidative cleavage of double 20 bonds and aromatic rings, oxidation of aromatic side chains, oxidative decarboxylation, and bisdecarboxylation), reactions involving replacement of hydrogen by oxygen (e.g., oxidation of methylene to carbonyl, oxidation of methylene to OH, CO 2 R, or OR, oxidation of arylmethanes, oxidation of ethers to carboxylic esters and related reactions, oxidation of aromatic hydrocarbons to quinones, oxidation of amines or nitro compounds to aldehydes, ketones, or dihalides, 25 oxidation of primary alcohols to carboxylic acids or carboxylic esters, oxidation of alkenes to aldehydes or ketones, oxidation of amines to nitroso compounds and hydroxylamines, oxidation of primary amines, oximes, azides, isocyanates, or notroso compounds, to nitro compounds, oxidation of thiols and other sulfur compounds to sulfonic acids), reactions in which oxygen is added to the subtrate (e.g., oxidation of alkynes to a-diketones, oxidation of tertiary amines to 30 amine oxides, oxidation of thioesters to sulfoxides and sulfones, and oxidation of carboxylic acids to peroxy acids), and oxidative coupling reactions (e.g., coupling involving carbanoins, dimerization of silyl enol ethers or of lithium enolates, and oxidation of thiols to disulfides).

WO 2004/016767 PCT/US2003/025984 , .. 77 102251 Exemplary reductive reactions include, for example, reactions involving replacement of oxygen by hydrogen (eg, reduction of carbonyl to methylene in aldehyds'and ketones, reduction of carboxylic acids to alcohols, reduction of aides to aniines, reduction of carboxlic esters to ethers, reduction of cyclic arihydridp to lactones and acid derivatives to 5 alcohols, reduction of carboxylic esteis to alcohols; reduction of carboxylic acids and esters to alkanes, complete reduction of epoxidds, reduttion of nitro compounds to amines, redu4ion 6f nitro compounds to hydroxylamines, reductionh.f nitroso'compounds and hydroxylamines to amines, rodudtion of oximes to primary amines or aziridines, reduction of azides to primary amines, reduction of nitrogen compounds, and reduction of sulfonyl halides and sulfonic acids to 10 thiols), removal of oxygen from the substrate (eig., reduction of amine oxides and azoxy compounds, reduction of sulfoxides and sulfones, reduction of hydroperoxides and peroxides, and reduction of aliphatic nitro compounds to oximes or nitriles), reductions that include cleavage (q.g., de-alkylation of amines and aides, redLction of azo, azoxy, and hydrazo compounds to amines, and reduction of disulfides to thiols), reductive couplic reactions (e.g., 15 bimolecular reduction of aldehydes and ketones to 1,2-diols, bimolecular reduction of aldehydes or ketones to alkenes, acyloin ester condensation, reduction of nitro to azoxy compounds, and reduction of nitro to azo compounds), and-reductions in which an organic substrate is both oxidized and reduced (e.g., the Cannizzaro reaction, the Tishchenko reaction, the Pummerer rearrangement, and the Willgerodt reaction). 20 (vi) Stereoselectivity 102261 The chiral nature of nucleic acids raises the possibility that nucleic acid-templated synthesis can proceed stereoselectively without the assistance of chiral groups beyond those present in the nucleic acid, thereby transferring not only sequence but also stereochemical information from the template to the product. Previous studies have demonstrated that the 25 chirality of nucleic acid templates can induce a preference for the template-directed ligation of (D)-nucleotides over (L)-nucleotides (Kozlov et al. (2000) ANGEW. CHEM. INT. ED. 39: 4292 4295; BoIli et al. (1997) A. CHEM. BIoLt 4: 309-320). [0227] During nucleic acid-templated synthesis it is possible to transfer the chirality of a nucleic acid template transfer unit, catalyst or a combination of the foregoing to reaction 30 products that do not resemble the nucleic acid backbone. In some embodiments, the reactive unit with a chiral center is associated with the template and the reactive unit associated with the WO 2004/016767 PCT/US2003/025984 -78 transfer unit is achiral, while in other embodiments, the transfer unit's reactive unit is chiral and the template's reactive unit is achiral. Alternatively, both reactive units can possess chiral centers. In each of these cases, the chirality of the template directs which of the chiral reactive unit's stereoisomers reacts preferentially (i.e., with a higherrate constant) with the other reactive 5 unit. f02281 Useful template architectures'include the H type,' E. type, Q type and T. type architecture. One or more template or transfer unit nucleotides may be replaced with non nucleotide linkers, however, replacement of the nucleotides nearest the reactive units may result in loss of stereoselectivity. Preferably, 5 or more consecutive aromatic nucleotides are adjacent 10 to the reactive units, and more preferably 6 or more consecutive aromatic nucleotides are adjacent to the reactive units. [02291 At high salt concentrations, double-stranded DNA sequences rich in (5-Me-C)G repeats can adopt a left-handed helix (Z-form) rather than the usual right-handed helix (B-form). During DNA-templated synthesis, template-transfer unit complexes in the Z-form cause 15 preferential reaction with one stereoisomer of a reactive unit, while template-transfer unit complexes in the B-form cause preferential reaction with the other stereoisomer of a reactive unit. Therefore, in some embodiments, a high concentration (e.g., at least 2.5 M, or at least 5 M) of a salt, such as, for example, sodium chloride (NaCl) or sodium sulfate (Na 2

SO

4 ) is used <uring DNA-templated synthesis. In other embodiments, the concentration of salt is low (e.g., not 20 greater than 100 mM) or is not present at all. The principles of DNA-templated stereospecific reactions are discussed in more detail in Example 6. (vii) Otherwise Incompatible Reactions [02301 It has been discovered that during nucleic acid-templated synthesis, oligonucleotides can simultaneously direct several different types of synthetic reactions within 25 the same solution, even though the reactants involved would be cross-reactive and therefore incompatible under traditional synthesis conditions (see, Example 7). As a result, nucleic acid templated synthesis permits one-pot diversification of synthetic library precursors into products of multiple reaction types. [02311 In one embodiment one or more templates associated with a single reactive unit 30 are exposed to two or more transfer units, each associated with a different reagent that is capable of reacting with the templates reactive unit. In other embodiments, one or more transfer units WO 2004/016767 PCT/US2003/025984 79' associated with a single reagent are exposed to'tw6 or more templates, each associated with a different reactive unit that is capable of reacting with'the reagent. Under the conditions of nucleic acid-templated synthesis, it is possible to haVe in a single solutiofi'multiple reactive units. (attached to the template and/or the transfer units) -that iii normal synthetic reactions would cross 5 react with' one anotlier:- The nucleic acid-terrplated chemistries described herein use very low concentritionstof reactants that because of concentration effects do not reach with one another: It is only when the reactants are brought together'via annealing of the oligonucleotide in the transfer unit to the template that their local concentrations are increased to permit a reaction occur. In some embodiments, a single accessory reagent (i.e., a reagent not linked.to a nucleic 10 acid or nucleic acid analog), such as, for example, a reducing agent, an oxidizing agent, or an activating agent, is added to the reaction. In other embodiments, no accessory reagent is added. In all cases, only the reactive units and reagents that are associated with complimentary oligonucleotides (i.e., that contain complimentary codon/anti-codon sequences) react to form a reaction product, demonstrating the ability of nucleic acid-templated synthesis to direct the 15 selective one-pot transformation of a single functional group into multiple distinct types of products. [02321 In another embodiment, templates and transfer units are provided as described above, but the template reactive units and transfer unit reagents react with one another using multiple different reaction types. In some embodiments, multiple different accessory reagents 20 are added to the reaction. Again, only reaction products resulting from complimentary template/transfer unit sequences are formed in appreciable amounts. 10233] In certain embodiments, multiple transfer unit reagents are capable of reacting with each template reactive unit, and some of the transfer unit reagents can cross-react with one another. Even in the presence of several different cross-reactive functional groups, only reaction 25 products resulting from complimentary template/transfer unit sequences are formed in appreciable amounts. These findings indicate that reactions of significantly different rates requiring a variety of accessory reagents can be directed by nucleic acid-templated synthesis in the same solution, even when both templates and reagents contain several different cross-reactive functional groups. The ability of nucleic acid templates to direct multiple reactions at 30 concentrations that exclude non-templated reactions from proceeding at appreciable rates mimics, in a single solution, a spatially separated set of reactions.

WO 2004/016767 PCT/US2003/025984 -80 (viii) Identification of New Chemical Reactions [02341 In another aspect of the intention, as illustrated in Figure 12, nucleic acid ternplated synthesis can be used to discover previously unknow n chemical reactions between two or more reactive units. To facilitate reaction discovery, multiple templates are synthesized, each 5 comprising a different reactive unit coupled to a different oligonucleotide. Each template oligonucleotide contains a coding region, whiich ideritifies the reactive unit attached to the template, and an annealing region. In some embodinients, other sequences are iricluded in the ternplate pligonucleotide, including, for example, PCR primer sites. Multiple transfer units are also prepared, each comprising a different reagent coupled to a different oligonucleotide. 10 [0235] To test for new bond-forming reactions, one or more templates are combined with one'or more transfer units under conditions that allow for hybridization of the transfer units to the templates. In some embodiments, non-DNA linked accessory molecules are added to the reaction, such as, for example, an activating agent or a catalyst. In other embodiments, reaction conditions, including, for example, reaction duration, temperature, solvent, and pH, are varied to 15 select reactions that proceed at different rates and under different conditions. [0236] The crude reaction mixture then is selected for particular reaction products. The reaction products preferably still are associated with their respective templates whose nucleotide sequence encodes the bond forming reactions that produced the reaction products. In some embodiments, the transfer unit is coupled to a capturable molecule, such as, for example, biotin. 20 Following creation and selection of the reaction products the associated templates can be selected by capturing the biotin by streptavidin. In one embodiment, the streptavidin is immobilized to a solid support, for example, by linkage to a magnetic bead. The selected templates then are amplified by PCR and subjected to DNA sequencing to determine the identities of the reactive unit and the reagent. In another embodiment, the reactions revealed by 25 the above approach are characterized in a non-DNA-templated format in both aqueous and organic solvents using traditional reaction analysis methods including, for example, thin-layer chromatography, NMR, HPLC, and mass spectroscopy. [02371 It is theoretically possible that some of the reactions discovered will require some aspect of the DNA template to proceed efficiently. However, the vast majority, if not all, of the 30 'reactions discovered in this system will take place in the absence of DNA template when performed at typical non-DNA-templated synthesis concentrations (e.g., about 0.1 M).

WO 2004/016767 PCT/1US2003/025984 Reactions discovered in this manner also are naturally weliL-suited for DNA-templated small molecule- library synthesis. An illustrative example of this embodiment appears in Example 12, describing the discovery of a new palladium"-mediated coupling'reaction between a terminal alkyne and a simple alkene. 5 (ix) Preparing Product Libraries [0238] A major prabtical difference between traditional and nucleic acid-templated library synthesis is the scale of each manipulation. Due to the amounts of material needed for screening and compound identification, traditional, 6orbinatorial syntheses typically proceed on the nmol-pml scale per library member.. In contrast, nucleic acid-templated library synthesis 10 can take place on the fnol-pmol scale because only minute quantities (e.g., about 10-20 mol) of each nucleic acid-linked synthetic molecule are needed for selection and PCR amplification. This vast difference in scale, combined with the:single-solution format of the nucleic acid templated libraries, simplifies significantly. the preparation of materials required for nucleic acid templated library syntheses. 15 [0239] Libraries can be produced via the template mediated syntheses described lierein. For example, the template may comprise one or rhore reactive units (for example, scaffold molecules). However, in each case the template contains a coding sequence that identifies the particular reactive unit associated with the oligonucleotide. A library of templates is initially subjected to one or more nucleic acid-templated bond formation reactions using reagents 20 attached to decoding oligonucleotides through a linker as described above. Depending upon the circumstances, the template library can be subjected to multiple iterations of bond formation reactions, wherein each intermediate product is purified before the subsequent round of reactions. In other circumstances, the intermediate products are not purified between reaction iterations. Preferably less than 20 bond forming reactions are required to create a library. In 25 other embodiments, less than 10 bond forming reaction steps are needed, and more preferably, between 3 and 7 steps are needed to create a full library. [0240] After the final round of nucleic acid-templated bond formation reactions has been performed accessory reagents can be added to protect exposed reactive functional groups on the reaction product, if necessary. In some embodiments, accessory reagents are added to initiate a 30 subsequent reaction with the reaction product, such as, for example, a cyclization reaction. The resulting library of reaction products attached to template oligonucleotides then are purified WU 2004/016767 PCT/US20031025984 - 82 and/oi selected as discussed herein. As would be appreciated by one skilled in this art, libraries ofsma1 niolecules or polymers can be synthesized using the principles discussed herein. [0241j Using similar approaches, it is possible to create a library of non-natural polymers from a library of template oligonucleotides that are not initially associated with a reactive unit. 5 In this case; the template encodes two or more codons which when annealed to corresponding anti-codons attached to monomer units bring together the Monomer units in a sequence specific manner. The transfer units then are allowed to contact the template under conditions that permit hybridization of the anti-codons on eachtransfer unit to the complementary codon on the template. Polymerization of the monomer units along the template then produces the polymer. 10 The polymerization may be step-by-step or may be essentially simultaneous with the chain being formed in one large reaction with one reaction between adjacent monomers leading to the attachment of the next monomer. In some embodiments, the functional group or groups of each monomer are protected, and must be depiotected prioi to polymerization. The newly synthesized polymer can then be cleaved from the anti-codons and the template, and selected for a desired 15 activity or characteristic, as described herein. DNA-templated polymer synthesis reactions are described in more detail in Example 9A and 9C. IV. SELECTION AND SCREENING [0242] Selection and/or screening for reaction products with desired activities (such as catalytic activity, binding affinity, or a particular effect in an activity assay) may be performed 20 according to any standard protocol. For example, affinity selections may be performed according to the principles used in library-based selection methods such as phage display, polysome display, and mRNA-fusion protein displayed peptides. Selection for catalytic activity may be performed by affinity selections on transition-state analog affinity columns (Baca et al. (1997) PRoc. NATL. ACAD. SCL USA 94(19): 10063-8) or by function-based selection schemes 25 (Pedersen et al (1998) PRoc. NATL. ACAD. Sci USA 95(18): 10523-8). Since minute quantities of DNA (10- 2 0 mol) can be amplified by PCR (Kramer et al. (1999) CURRENT PROTOCOLS IN MOLECULAR BIOLOGY (ed. Ausubel, F. M) 15, 1-15.3, Wiley), these selections can be conducted on a scale ten or more orders of magnitude less than that required for reaction analysis by current methods, making a truly broad search both economical and efficient.

WO 2004/016767 PCT/US2003/025984 -83 (i) Selection for Binding to Target Molecule [02431 The templates and reaction products can be selected (or screened) for binding to a target njolecule. In this context, selection or partitioning means any process whereby a library member bound to a target molecule is separated from library members not bound to target 5 molecules.' Selection can be accomplished by various methods known in the art. (02441 The templates of the present invention contain a built-in function for direct selection and amplification. In most applications, binding to a target molecule preferably is selective,'such that the template and the resulting reaction product bind preferentially with a specific target molecule, perhaps preventing or inducing a specific biological effect. Ultimately, .0 a binding molecule identified using the present invention may be useful as a therapeutic and/or diagnostic agent. Once the selection is complete, the selected templates optionally can be amplified and sequenced. The selected reaction products, if present in sufficient quantity, can be separated from the templates, purified (e.g., by HPLC, column chromatography, or other chromatographic method), and further characterized. 5 (ii) Target Molecules [0245] , Binding assays provide a rapid means for isolating and identifying reaction products that bind to, for example, a surface (such as metal, plastic, composite, glass, ceramics, rubber, skin, or tissue); a polymer; a catalyst; or a target biomolecule such as a nucleic acid, a protein (including enzymes, receptors, antibodies, and glycoproteins), a signal molecule (such as 0 cAMP, inositol triphosphate, peptides, or prostaglandins), a carbohydrate, or a lipid. Binding assays can be advantageously combined with activity assays for the effect of a reaction product on a function of a target molecule. [0246] The selection strategy can be carried out to allow selection against almost any target. Importantly, the selection strategy does not require any detailed structural information 25 about the target molecule or about the molecules in the libraries. The entire process is driven by the binding affinity involved in the specific recognition and binding of the molecules in the library to a given target. Examples of various selection procedures are described below. [02471 The libraries of the present invention can contain molecules that could potentially bind to any known or unknown target. The binding region of a target molecule could include a 0 catalytic site of an enzyme, a binding pocket on a receptor (for example, a G-protein coupled WO 2004/016767 PCT/US2003/025984 84 receptor), a protein surface area involved in a prbtejn-protein or protein-nucleic acid interaction (preferably a hot-sp6t region), or a specific site on DNA'(such as the major groove). Th riatural function of the -target could be stimulated (agonized), reduced (antagonized), unaffected, or completely changed by the binding of the reacti o n product. This will depend on the precise 5 binding mdde and thd particular binding site the reaction product occupies on the target, [02481 Functioxgal sites (such as protein-protein interaction or catalytic sites) on proteins often are more prone to bind molecules than are other more neutral surface areas on a protein. In addition, these functional sites normally contain a smaller region that seems to be primarily responsible for the binding energy: the so-called "hot-spot regions" (Wells, et a]. (1993) RkaFzNT 10 PROG. HORMONE REs. 48: 253- 262). This phenomenon facilitates selection for molecules affecting the biological function of a certain target.

[02491 The linkage between the template'mol eule and reaction product allows rapid identification of binding molecules using various selection strategies. This invention broadly permits identifying binding molecules for any known target molecule. In addition, novel 15 unknown targets can be discovered by isolating binding molecules against unknown antigens (epitopes) and using these binding molecules for identification and validation. In another preferred embodiment, the target molecule is designed to mimic a transition state of a chemical reaction; one or more reaction products resulting from the selection may stabilize the transition state and catalyze the chemical reaction. 20 (iii) Binding Assays [0250] The template-directed synthesis of the invention permits selection procedures analogous to other display methods such as phage display (Smith (1985) SCIENCE 228: 1315 1317). Phage display selection has been used successfully on peptides (Wells et al. (1992) CURR. OP. STRUCT. BIOL. 2: 597-604), proteins (Marks et al (1992) J. BioL. CHEM. 267: 16007 25 16010) and antibodies (Winter et aL (1994) ANNU. REV. IMMUNOL. 12: 433-455). Similar selection procedures also are exploited for other types of display systems such as ribosome display Mattheakis et aL (1994) PRoc. NATL. ACAD. SC. 91: 9022-9026) and mRNA display (Roberts, et al (1997) PROC. NATL. ACAD. SCL 94:12297-302). The libraries of the present invention, however, allow direct selection of target-specific molecules without requiring 30 traditional ribosome-mediated translation. The present invention also allows the display of small molecules which have not previously been synthesized directly from a nucleic acid template.

WO 2004/016767 PCT/US2003/025984 -85 102511 Selection of binding molecules from a library can be performed in-any format to identify optimal binding molecules. Binding selections typically involve immobilizing the, desired target molecule, adding a library of potential binders, and removing non-binders by washing. When the molecules showing low affinity for an immobilized target are washed away, 5 the molecules with a stronger affinity generally remain attached to the target. The enriched population remaining bound to the target after stringent washing is preferably eluted with, for example, acid, chaotropic salts, heat, competitive elution with a known ligand or-by proteolytic release of the target and/or of t9ihplate molecules. The eluted templates are suitable for PCR, leading to many orders of amplification, whereby essentially each selected template becomes 10 available at a greatly increased copy number for cloning, sequencing, and/or further enrichment or diversification. [02521 In a binding assay, when the concentration of ligand is much less than that of the target (as it would be during the selection of a DNA-femplated library), the fraction of ligand bound to target is determined by the effective concentration of the target protein (see, Figure 15 10). The fraction of ligand bound to target is a sigmoidal function of the concentration of target, with the midpoint (50% bound) at [target] = KAd of the ligand-target complex. This relationship indicates that the stringency of a specific selection - the minimum ligand affinity required to remain bound to the target during the selection - is determined by the target concentration. Therefore, selection stringency is controllable by varying the effective concentration of target. 20 [02531 The target molecule (peptide, protein, DNA or other antigen) can be immobilized on a solid support, for example, a container wall, a wall of a microtiter plate well. The library preferably is dissolved in aqueous binding buffer in one pot and equilibrated in the presence of immobilized target molecule. Non-binders are washed away with buffer. Those molecules that may be binding to the target molecule through their attached DNA templates rather than through 25 their synthetic moieties can be eliminated by washing the bound library with unftnctionalized templates lacking PCR primer binding sites. Remaining bound library members then can be eluted, for example, by denaturation. [02541 Alternatively, the target molecule can be immobilized on beads, particularly if there is doubt that the target molecule will adsorb sufficiently to a container wall, as may be the 30 case for an unfolded target eluted from an SDS-PAGE gel. The derivatized beads can then be used to separate high-affinity library members from nonbinders by simply sedimenting the beads V LU4/M110 O PCT/US2003/025984 '- 86-' in a benchtop centrifuge. Alternatively, the beads can be used to make an affinity column. In such cases, the library is passed through'the column one'or more times to pemit-binding. The column then is washed to remove nonliinding library roernbers. Magnetic beads are essentially a variant'on the above; the target is attached to magnetic b6ads which are then used in the 5 selection. [02551 There are many reactive natrices avail ble for immobilizing the target molecule, including matrices bearing -NIH 2 groups or -SH 'groups. The target molecule can be immobilized by conjugation with NHS ester or maleimide groups coyalently linked to Sepharose bea& and the integrity of known properties of the target molecule can be verified. Activated beads are, 0 available with attachment sites for -NH 2 or -COOH groups (which can be used for coupling). Alternatively, the target molecule is blotted onto nitrocellulose or PVDF. When using a blotting strategy, the blot should be blocked (e.g., with BSA or similar protein)'after inmobilization of the target to pr6yent nonspecific binding of library members to the blot. [0256] Library members that bind a target molecule can be released by denaturatioh, 5 acid, or chaotropic salts. Alternatively, elution conditions can be more specific to reduce i background or to select for a desired specificity. Elution can be accomplished using proteolysis to cleave a linker between the target molecule and the immobilizing surface or between the' reaction product and the template. Also, elution cin be accomplished by competition with a known competitive ligand for the target molecule. Alternatively, a PCR reaction can be performed directly in the presence of the washed target molecules at the end of the selection procedure. Thus, the binding molecules need not be elutable from the target to be selectable since only the template is needed for farther amplification or cloning, not the reaction product itself Indeed, some target molecules bind the most avid ligands so tightly that elution would be difficult. [02571 To select for a molecule that binds a protein expressible on a cell surface, such as an ion channel or a transmembrane receptor, the cells themselves can be used as the selection agent. The library preferably is first exposed to cells not expressing the target molecule on their surfaces to remove library members that bind specifically or non specifically to other cell surface epitopes. Alternatively, cells lacking the target molecule are present in large excess in the selection process and separable (by fluoresdence-activated cell sorting (FACS), for example) from cells bearing the target molecule. In either method, cells bearing the target molecule then WO 2004/016767 PCT/US2003/025984 -87 are used to isolate library members bearing the target molecule (e.g., by sedimenting the cells or by FACS Sorting). For example, a recombinant DNA encoding the target molecule can be introduced into a cell line; library members that bind the tansformed cells but not the untransformed cells are enriched for target molecule binders. This approach is also called 5 subtraction selection and has successfully been used for phage display on antibody libraries (Hoogenboom et a. (1998) IMMUNOTECH 4:1- 20). [0258] A selection procedure can also involve selection for binding to cell surface reqeptors that are internalized so that the.receptor together with the selected binding molecule passes into the cytoplasm, nucleus, or other cellular compartment, such as the Golgi or 10 lysosomes. Depending on the dissociation rate constant for specific selected binding molecules, these molecules may localize primarily within the intracellular compartments. Internalized library members can be distinguished from molecules attached to the cell surface by washing the cells, preferably with a denaturant. More preferably, standard subcellular fractionation techniques are used to isolate the selected library members in a desired subcellular compartment. 15 [0259] An alternative selection protocol also includes a known, weak ligand affixed to each member of the library. The known ligand guides the selection by interacting with a defined part of the target molecule and focuses the selection on molecules that bind to the same region, providing a cooperative effect. This can be particularly useful for increasing the affinity of a ligand with a desired biological function but with too low a potency. 20 10260J Other methods for selection or partitioning are also available for use with the present invention. These include, for example: immunoprecipitation (direct or indirect) where the target molecule is captured together with library members; mobility shift assays in agarose or polyacrylamide gels, where the selected library members migrate with the target molecule in a gel; cesium chloride gradient centrifugation to isolate the target molecule with library members; 25 mass spectroscopy to identify target molecules labeled with library members. In general, any method where the library member/ target molecule complex can be separated from library members not bound to the target is useful. [02611 The selection process is well suited for optimizations, where the selection steps are made in series, starting with the selection of binding molecules and ending with an optimized 30 binding molecule. The procedures in each step can be automated using various robotic systems. Thus, the invention permits supplying a suitable library and target molecule to a fully automatic WO 2004/016767 PCT/US2003/025984 , 1 -88 system which finally generates an optimized bindingmolecule. Under ideal conditions, this process should run without any requirenientfb external work outside the robotic system during the entire procedure. [0262] The selection methods of the present invention can be conibined with secondary 5 selection or screening to identify reaction products'capable of modifying target molecuib function upon bihding., Thus, the methodsdescribed herein car be employed to isolate or. produce binding molecules that bind to and niodify the function of any protein or nucleic acid. For example, nucleic acid-templated chemistry can be used to identify, isolate, or produce binding molecules (1) affecting catalytic activity of target enzymes by inhibiting catalysis or, 10 modifying substrate binding; (2) affecting the functionality of protein receptors, by inhibiting binding to receptors-or by'modifying the specificity of binding to receptors; (3) affecting the formation of protein multirners by disrupting the quateitary structure of protein subynits; or (4) modifying transport-properties of a protein by disrupting transport of small molecules or ions [02631 Functional assays can be included in the selection process. For example,,after 15 selecting for binding activity, selected library members can be directly tested for a desired, functional effect, such as an effect on cell signaling. This can, for example, be performed via FACS methodologies. [02641 The binding molecules of the invention can be selected for other properties in addition to binding. For example, to select for stability of binding interactions in a desired 20 working environment. If stability in the presence of a certain protease is desired, that protease can be part of the buffer medium used during selection. Similarly, the selection can be performed in serum or cell extracts or in any type of medium, aqueous or organic. Conditions that disrupt or degrade the template should however be avoided to allow subsequent amplification. 25 (iv) Other Selections [0265] Selections for other desired properties, such as catalytic or other functional activities, can also be performed. Generally, the selection should be designed such that library members with the desired activity are isolatable on that basis from other library members. For example, library members can be screened for the ability to fold or otherwise significantly 30 change conformation in the presence of a target molecule, such as a metal ion, or under particular pH or salinity conditions. The folded library members can be isolated by performing WO 2004/016767 PCT/US2003/025984 -89 non-deraturing gel electrophoresis under the conditions of interest. The folded library members migrate to a different position in the gel-andcin avbsequently be extracted from the gel arid isolated. . 102661, Similarly, reaction products that fludresce in the presence of specific ligands may 5 be selected by FACS based sorting of translated pblymers linked through their DNA templates to beads. Those bbads that fluoresce in the presence, but not in the absence, of'the target ligand are isolated andcharacterized. Useful beads with a homogenous population of nucleic acid templates'on any bead can be prepared using the splitpool synthesis technique on the bead, such that each bead is exposed to only a single nucleotide sequence. Alternatively, a different aiti 10 template (each complementary to only a single, different template) can by synthesized on beads using a split-pool technique, and then can anneal to capture a solution-phase library. [0267] .Biotin-terminated biopolyniers can be selected for the actual catalysis of bond breaking reactions by passing these biopolymers over a resin linked through a substrate to avidin (Figure 1A). Those biopolymers that catalyze substrate bleavage self-elute from a column 15 charged with this resin. Similarly, biotin-terminated biopolymers can be selected for the catalysis of bond-forming reactions (see, Figure 11B). One substrate is linked to resin and the second substrate is linked to avidin. Biopolymers that catalyze bond formation between the substrates are selected by their ability to react the'substrates together, resulting in attachment of the biopolymer to the resin. 20 [0268] . Library members can also be selected for their catalytic effects on synthesis of a polymer to which the template is or becomes attached. For example, the library member may influence the selection of monomer units to be polymerized as well as how the polymerization reaction takes place (e.g., stereochemistry, tacticity, activity). The synthesized polymers can be selected for specific properties, such as, molecular weight, density, hydrophobicity, tacticity, 25 stereoselectivity, using standard techniques, such as, electrophoresis, gel filtration, centrifugal sedimentation, or partitioning into solvents of different hydrophobicities. The attached template that directed the synthesis of the polymer can then be identified. [02691 Library members that catalyze virtually any reaction causing bond formation between two substrate molecules or resulting in bond breakage into two product molecules can 30 be selected using the schemes proposed in Figures 12 and 13. To select for bond forming catalysts (for example, hetero Diels-Alder, Heck coupling, aldol reaction, or olefin metathesis WO 2004/016767 PCT/US2003/025984 -90 catalysts), library members are covalently linked to one substrate through their 5'.amino or thiol termini. The other substrate of the reaction is synthesized as a derivative linked to biotin. 'When dilute solutions of library-substrate conjugate are combined with the substrate-biotin conjugate, those library members that catalyze bond foration cause the biotin group to become covalently 5 attached to .themselves. Active bond forming catalysts can then be separated from inactive library members by capturing the former with immobilized streptavidin and washing atway inactive library members (Figure 12). 102701 In an analogous manner, library menibers that catalyze bond cleavage reactions such as retro-aldol reactions, amide hydrolysis, elimination reactions, or olefin dihydroxylation 10 followed by periodate cleavage can be selected. In this case, library members are e6valently linked to biotinylated substrates such that the bond breakage reaction causes the disconnection of the biotin moiety from the library members (Figure 13). Upon incubation under reaction conditions, active catalysts, but not inactive library niembers, induce the loss of their biotin groups. Streptavidin-linked beads can then be used to capture inactive polymers, while active 15 catalysts are able to be eluted from the beads. Related bond formation and bond cleavage selections have been used successfully in catalytic RNA and DNA evolution (Jischke et aL (2000) CURR. OPIN. CHEM. BIot 4: 257-62) Although these selections do not explicitly select for multiple turnover catalysis, RNAs and DNAs selected in this manner have in general proven to be multiple turnover catalysts when separated from their substrate moieties (Jhschke et al. 20 (2000) CURR. OPIN. CHEM. BIOL. 4: 257-62; Jaeger et aL (1999) PROC. NATL. ACAD. Sci. USA 96: 14712-7; Bartel et al (1993) SCIENCE 261: 1411-8; Sen et aL (1998) CURR. OPIN. CHEM. BIOL. 2: 680-7). [0271] In addition to simply evolving active catalysts, the in vitro selections described above are used to evolve non-natural polymer libraries in powerful directions difficult to achieve 25 using other catalyst discovery approaches. Substrate specificity among catalysts can be selected by selecting for active catalysts in the presence of the desired substrate and then selecting for inactive catalysts in the presence of one or more undesired substrates. If the desired and undesired substrates differ by their configuration at one or more stereocenters, enantioselective or diastercoselective catalysts can emerge from rounds of selection. Similarly, metal selectivity 30 can be evolved by selecting for active catalysts in the presence of desired metals and selecting for inactive catalysts in the presence of undesired metals. Conversely, catalysts with broad WO 2004/016767 PCT/US2003/025984 I , - -91-' substrate tolerance can be evolved by varying substrate structures between successive rounds of I . F selection. (v) Iterative Selection [0272] Iterating a selection by loading eluant from a first selection into a second selection 5 multiplies the net enrichment. No intervening amplification of template is required. For example, a selection fdr binding to carbonic anhydrase beads permitted a 330-fold enrichment of a ligand. .Application of the eluant directly to fresh carbonic anhydrase beads (see, Example 11) enriched the template encoding the carbonic anhydrase ligand >10,000-fold Where the selection was repeated a third time, a 5,000,000-fold net enrichment'of the ligand was observed. This 10 result indicates that iterating library selections' can lead to very large enrichments of desired molecules. In certain embodiments, a first rpund of selection provides at least a 50-fold increase in the number of binding ligands. Preferably, the increse in enrichmehts is over 100-fold, more preferably over 1,000 fold, and even more preferably over 100,000-fold. Subsequent rounds of selection may further increase the enrichment 100-fold over the original library, preferably 15 1,000-fold, more preferably over 100,000-fold, and most preferably over 1,000,000-fold. [0273] Alternatively, following PCR amplification of DNA templates encoding selected synthetic molecules, additional rounds of translation, selection, and amplification can be conducted to enrich the library for high affinity binders. 'The stringency of the selection is gradually increased by increasing the salt concentration of the binding and washing buffers, 20 decreasing the duration of binding, elevating the binding and washing temperatures, and increasing the concentration of washing additives such as template DNA or unrelated proteins. [0274] Importantly, in vitro selections can also select for specificity in addition to binding affinity. Library screening methods for binding specificity typically require duplicating the entire screen for each target or non-target of interest. In contrast, selections for specificity 25 can be performed in a single experiment by selecting for target binding as well as for the inability to bind one or more non-targets. Thus, the library can be pre-depleted by removing library members that bind to a non-target. Alternatively, or in addition, selection for binding to the target molecule can be performed in the presence of an excess of one or more non-targets, as described in Example 11. To maximize specificity, the non-target can be a homologous 30 molecule. If the target molecule is a protein, appropriate non-target proteins include, for example, a generally promiscuous protein such as an albumin. If the binding assay is designed WO 2004/016767 PCT/UJS20031025984 -92 to target only a specific portion of a target molecule, the non-target can be a variation on the molecule in which that portion has been changed or removed. (W) A nplificadon and Sequencing [0275 - Once all rounds of selection are complete, the templates which are, or formerly 5 were, associated with the selected reaction product preferably'are amplified using. any suitable technique to facilitate sequencing or other subsequent'manipulation of the templates. Natural oligonucleotides can be amplified by any state of the art method. These methods include, for example, polymerase chain reaction (PCR); nucleic acid sequence-based amplification (see, for example, Compton (1991) NATURE 350: 91-92), amplified anti-sense RNA (see, for example, 10 van Gelder et a. (1988) PROC. NATL. ACAD. Sci. USA 85: 77652-77656); self-sustained sequence replication systems (Gnatelli et al. (1990) PROC. NATL. ACAD. SC. USA 87: 1874 1878); polymerase-independent amplification (see, for example, Schmidt et al (1997) NUCLEIc ACIDs RES. 25: 4797-4802, and in vivo amplification of plasmids carrying cloned DNA fragments. Descriptions of PCR methods are found, for example, in Saiki et al (1985) SCIENCE 15 230: 1350-1354; Scharf et al (1986) SCIENCE 233: 1076-1078; and in U.S. Patent No. 4,683,202. Ligase-mediated amplification methods such as Ligase Chain Reaction (LCR) may also be used. In general, any means allowing faithful, efficient amplification of selected nucleic acid sequences can be employed in the niethod of the present invention. It is preferable, although not necessary, that the proportionate representations of the sequences after amplification reflect the 20 relative proportions of sequences in the mixture before amplification. [02761 For non-natural nucleotides the choices of efficient amplification procedures are fewer. As non-natural nucleotides can be incorporated by certain enzymes including polymerases it will be possible to perform manual polymerase chain reaction by adding the polymerase during each extension cycle. 25 [02771 For oligonucleotides containing nucleotide analogs, fewer methods for amplification exist. One may use non-enzyme mediated amplification schemes (Schmidt et al. (1997) NUCLEIC ACIDS REs. 25: 4797-4802). For backbone-modified oligonucleotides such as PNA and LNA, this amplification method may be used. Alternatively, standard PCR can be used to amplify a DNA from a PNA or LNA oligonucleotide template. Before or during 30 amplification the templates or complementing templates may be mutagenized or recombined in order to create an evolved library for the next round of selection or screening.

WO 2004/016767 PCT/US2003/025984 - 93, (vii) Sequence Determination ' 102781 Sequencing can be done by' a stafidard dideoxy chain termination method, or by chemical sequencing, for example, using the Maxam-Gilbert sequencing procedure.

Alternatively,. the sequence of the template (or, if a long template is used, the variable portion(s) 5 thereof) can be determined by hybridization to a chip (see, Example 12). For'example, a single stranded templiate molecule associated with-a detectable nioiety such as a fluorescent moiety iI exposed to a chip beating a large number of clpnal populations of single-stranded nucleic acids, or nucleic acid analogs'of known sequences each 6lonal population being present at a particular addressable location on the chip. The template sequences are permitted to anneal to the'chip, 10 sequences. The position of the detectable moieties on the chip then is determined. Based upon the location of the detectable moiety and th&inimobilized sequence at that location, the sequence of the template can be determined. It is contemplated that large numbers of such oligonucleotides can be immobilized in an array on a chip or other solid'support. (viii) Diversification 15 [0279] Inventive libraries can be evolved by introducing mutations at the DNA level, for example, using error-prone PCR (Cadwell et al. (1992) PCR METHODS APPL. 2: 28) or by subjecting the DNA to in vitro homologous recombination (Stemmer (1994) PROc. NATL. ACAD. Sci. USA 91: 10747; Stemmer (1994) NATuR.E 370: 389). [0280] Small molecule evolution using mutation and recombination offers two potential 20 advantages over simple enrichment. If the total diversity of the library is much less than the number of molecules made (typically 1012 to 101), every possible library member is present at the start of the selection. In this case, diversification is still useful because selection conditions can change as rounds of evolution progress. For example, later rounds of selection can be conducted under higher stringencies and can involve counterselections against binding to non 25 target molecules. Diversification gives library members that have been discarded during earlier rounds of selection the chance to reappear in later rounds under altered selection conditions in which their fitness relative to other members may be greater. In addition, it is quite possible to generate a synthetic library that has a theoretical diversity greater than 10is molecules. In this case, diversification allows molecules that never existed in the original library to emerge in later 30 rounds of selections on the basis of their similarity to selected molecules, similar to the way in WO 2004/016767 PCTIUS2003/025984 -94 which protein evolution searches the vastness of protein sequence space one small subset at a time. (iii)(a) Error-prone PCR 102811 Random point mutagenesis is perfonned by conducting the PCR amplification 5 step under error-prone PCR (Cadwell et al. (1992) PCR METHODS APPLIC. 2: 28.-33) conditions.

Because the genetic code of these molecules are written to assign related codons to related chemical groups, similar to the way that the natural protein genetic code is constructed, random point mutations in the templates encoding selected molecules will diversify progriy towards chemically related analogs. Because error-prone PCR is inherently less efficient than normal 10 PCR, error-prone PCR diversification is preferably conducted with only natural dATP, dTTP, dCTP, and dGTP and using primers that lack chemical handles or biotin groups. (viii)(b) Recombination [02821 Libraries may be diversified using recombination. For example, templates to be recombined may have the structure shown in Figure 14, in which codons are separated by five 15 base non-palindromic restriction endonuclease cleavage sites such as those cleaved by AvaII (G/GWCC, W=A or T), Sau96I (G/GNCC, N=A, G, T, or C), Ddel (C/TNAG), or HinF (G/ANTC). Following selections,. templates encoding desired molecules are enzymatically digested with these commercially available restriction enzymes. The digested fragments ihen are recombined into intact templates with T4 DNA ligase. Because the restriction sites separating 20 codons are nonpalindromic, template fragments can only reassemble to form intact recombined templates (Figure 14). DNA-templated translation of recombined templates provides recombined small molecules. In this way, functional groups between synthetic small molecules with desired activities are recombined in a manner analogous to the recombination of amino acid residues between proteins in Nature. It is well appreciated that recombination explores the 25 sequence space of a molecule much more efficiently than point mutagenesis alone (Minshull et aL. (1999) CURR. OPIN. CHEM. BIOL. 3: 284-90; Bogarad et a. (1999) PROc. NATL. ACAD. SC. USA 96: 2591-5; Stemmer NATURE 370: 389-391). [0283] A preferred method of diversifying library members is through nonhomologous random recombination, as described, for example, in WO 02/074978; US Patent Application 30 Publication No. 2003-0027180-Al; and Bittker et al. (2002) NA TURE BIOTECH. 20(10): 1024-9.

WO 2004/016767 PCTIUS2003/025984 -95-. (iiv)(c) Random Cassette Mutagenesis [02841 Random cassette mutageiesis is useful to create a diversified library from a fixed starting sequence. Thus, such a method 'can be used, for example, after a library has been subjected to selection and one or more library members have been isolated and sequenced. 5 Generally,. a library of oligonucleotides with variations on the starting sequence is generated by traditional chemical synthesis, error-prone PCR, or other methods. For example, a library of oligonucleotides can be generated in which, for each nucleotide position in a codon, the inuejeotide has a 90% probability of being identicalto the starting sequence at that position, and a 10% probability of being different. The oligonucleotides can be complete templates when 10 synthesized, or can be fragments that are. subsequently ligated with other oligonibcleotides to form a diverse library of templates, V. USES [02851 The methods and compositions of the present invention represent new ways to generate molecules with desired properties. This approach marries extremely powerful genetic 15 methods, which molecular biologists have taken advantage of for decades, with the flexibility and power of organic chemistry. The ability to prepare, amplify, and evolve unnatural polymers by geneic selection may lead to new classes of catalysts that possess activity, bioavailability, stability, fluorescence, photolability, or other properties that are difficult or impossible to achieve using the limited set of building blocks found in proteins and nucleic acids. Similarly, 20 developing new systems for preparing, amplifying, and evolving small molecules by iterated cycles of mutation and selection may lead to the isolation of novel ligands or drugs with properties superior to those isolated by slower traditional drug discovery methods. [02861 For example, unnatural biopolymers useful as artificial receptors to selectively bind molecules or as catalysts for chemical reactions can be isolated. Characterization of these 25 molecules would provide important insight into the ability of polycarbamates, polyureas, polyesters, polycarbonates, polypeptides with unnatural side chain and stereochemistries, or other unnatural polymers to form secondary or tertiary structures with binding or catalytic properties. [0287] The present invention further allows the discovery of new chemical reactions. 30 The field of chemistry is continually being transformed by the discovery of new chemical reactions providing access to previously inaccessible molecules, allowing for expedited WO 2004/016767 PCT/US2003/025984 -96 syntheses, and revealing new chemical principles: Guided by predictions of reactivity based on literature precedent, chemists typically search for a new reaction to overcome a particulat shortcoming in current synthetic methodology. Until now, it has not been feasible to conduct a broad, nbn-biased search for chemical reactivity i1 which a large number'of diverse reactants are 5 simultaneously evaluated for their ability to react with one another under many different, conditions. Both'.the amount of material required for executing thousands of diverse reactions' and the difficulty of analyzing the outcome of.sudh an6xperiment makes this goal intractable using current reaction discovery approaches. A broad, non-biased search for chemical reactivity is appealing because it is not limited by conventional wisdom or by our ability to predict 10 functional group reactivity [0288] The inventive method of discovering new chemical reactions and chemical reactivity has several advantages over existing methods. For example, several groups have developed high -throughput screens to test the efficiency of a particular reaction under a variety of conditions (Kuntz et al. (1999) CURR. OPn. CHEM. BIOL. 3: 313-319; Francis et al. (1998) 15 CURR. OPrN. CHEM. BIOL. 2: 422-428; Pawlas et a. (2002) J. AM. CHEM. Soc. 124: 3669-3679; Lober el at. (2001) J. AM. CHEM. Soc. 123: 4366-4367; Evans et al. (2002) CuRR. OPN. CHEM. BIOL. 6: 333-338; Taylor et at. (1998) SCIENCE 280: 267-270; and Stambuli et al. (2001) J. AM. CHEM. SOC. 123: 2677-2678); however, the screens are limited to a small set of reaction types, Reactions have been analyzed in a high-throughput manner using fluorescence spectroscopy, 20 colorimetric assay, thermographic analysis, and traditional chromatography (Dahren et at. (2001) SYNTHESIS-STUTTGART 1431-1449 and Wennemers (2001) CoMBINATORIAL CHEMISTRY & HIGH THROUGHPUT SCREENING 4: 273-285). Most high-throughput screens for chemical reactivity are useful for only a small set of reaction types because the screen depends on a particular property of the reaction such as the disappearance of an amine or the production of 25 protons. As a result, high throughput screening methods can be useful for discovering catalysts for a known or anticipated reason, but are poorly suited to discover novel reactivity different from a reaction of interest. A non-biased search for chemical reactions would examine a broad range of both reaction conditions and reactants in a highly efficient manner that is practical on the scale of thousands of different reactions. The inventive method of discovering chemical 30 reactions offers a much greater chance of discovering unexpected and unprecedented reactivity that may lead to new insights into reactivity and to useful new reactions for chemical synthesis.

WO 2004/016767 PCT/US2003/025984 -97 [0Z891 Discovering new reactions from very large and diverse collections of reactants. afnd conditions entails (1) a general assay for reactivity that.does not depend on a particular substrate or product, and (2) increasing the overall efficiency of assaying reactions suchthat both reaction condition space and reactant space can be searched extensively. For example, 5 researchers evolving catalytic nucleic acids routinely select for bond formation catalysts by attaching one reactant to the pool of evolving nucleic acids and linking another reactant to a handle that can be easily immobilized such as biotii (Wilson et aL (1999) ANNU. REv. BIOCHEM. 68: 611-647; Jaschke (2001) CuRR. OPIN. STRUCT. B1OL. 11: 321-326; Jaschke et al. (2000) CURR. OPIN. CHEM. BIOL. 4: 257-262; Jaschke (2001) BIOL. CHiM. 382: 1321-1325). Active 10 nucleic acids become linked to the handle and are separated from the inactive sequences. Because this type of selection does not depend on the consumption or generation of a specific substrate or product, the scope of reactants that can' be tested in this type of selection is much larger than the scope of reactants that can be evaluated in current reactivity screens. 10290] Nucleic acid-templated synthesis provides a way to use bond formation selections 15 to discover new chemical reactivity independent of nucleic acid catalysis (Gartner et al. (2002) ANCEW. CHEM. INT. ED. 41: 1796-1800; Gartner et at (2001) supra). Nucleic acid templates can direct a wide variety of chemical reactions in a highly sequence-specific manner without any obvious requirements for reaction geometry. By attaching reactants to appropriately designed nucleic acid sequences, it becomes possible to test thousands of unprecedented reactions in a 20 single pot with individual sequences encoding each reaction. Pools of nucleic acid-linked reactants would be truly selected (not simply screened) for covalent bond formation with members of a second nucleic acid-linked reactant pool. PCR amplification and DNA sequencing would reveal which combinations of reactants successfully undergo bond formation. [0291] In certain embodiments, the searchable reactions are those transformations that 25 can occur in aqueous or substantially aqueous medium. In other embodiments, the searchable reactions are limited to those that do not degrade nucleic acids rapidly. The known chemical robustness of DNA suggests that a wide range of reaction conditions spanning different temperatures, pH ranges, and additives such as transition metals are compatible with the proposed approach. A DNA-templated Heck reaction demonstrates that transition metal 30 catalyzed reactions are viable in a DNA-templated format, consistent with extensive evidence (Patolsky et al (2002) J. AM. CHEM. Soc. 124: 770-772; Weizman et al. (2002) J. AM. CHEM.

WO 2004/016767 PCT/US2003/025984 -98 Soc. 124: 1568-1569; Gartner et al. (2002) ANGE CHEM. INT. ED. 41: 1796-1800; Czlapinski et al. (2001) J. AM.CHEM. Soc. 123: 8618-869; Jolmilin eta. (1998) J. AM. CLEM. Soc. 120: 9724-9725; Bashkii etal. (1994) J. AM. CHEM. Soe 116:.-5981 1 5982; MAgda et a. (1994) J. AM. CHEM. Soc. 116; 7439-7440; and Dandlik&r et al. (1997) SCIENCE 275: 1465-1468) that 5 DNA is compatible with many transition metal coinplexes, including those containing pd, Ni, Mn, Pt, Ru, Os; Cu, Eu, and Rh. Further, the'rapid increase in the number of known water compatible organic reactions (Li et al. Organii reaction in aqueous media (Wiley and Sons, New Yoi-k, 1997) and the inherent benefits of working in aqueous solvents suggests that, water is a rich medium for discovering new reactions; Reactiois discovered in this effort may be of 10 general utility when performed in a standard non-nucleic acid-templated mode, and are also natural candidates for use in generating nucleic acid-teinplated synthetic libraries. 102921 Nucleic acid-templated chemistry is combined with in vitro selection and PCR amplification in certain embodiments to efficiently search for novel bond-fnrming reactions independent-of reactant structures. The ability to select directly for covalent bond formation, the 15 minute scale required for analysis, and compatibility of nucleic acids with a wide variety of reaction conditions may permit the first search for unprecedented reactivity that can examine thousands of combinations of reactants aild reaction conditions in one or several experiments. [02931 The reaction generality and distance independence of DNA-templated synthesis allows for a system for discovering new chemical reactions by selection. DNA-linked reactants 20 (i.e., templates and/or transfer units) suitable for in vitro selection for bond formation exist in one or two forms designated pool A and pool B in Figure 9. Each reactant in pool B contains a functional group being tested linked to a short segment of biotinylated DNA (a coding region) encoding that functional group. Each reactant in pool A contains a functional group being tested, a corresponding coding region, and an "annealing region" or anti-codon that complements one of 25 the pool B coding regions. Each functional group in pool A is linked to one of every possible annealing region. This arrangement allows any functional group in pool A to join any functional group in pool B on the same DNA duplex, providing the opportunity for DNA-templated bond formation if the reactants are mutually reactive. Generating these two pools of DNA-linked reactants in a format suitable for in vitro selection for bond formation requires the development 30 of methods to efficiently assemble a small molecule reactant, a coding region, and in the case of pool A, a library of annealing regions.

w zuu4/t16767 PCT/US2003/025984 -99 [02941 The inventive system is particularly useful for the identification of small rnolecule/target binding pairs. For instance, inventive DNA-templated small molecule libraries may be contacted with other solution dr solid-phase librahres of potential target compounds such that small molecules within the inventive library that bind or interact with one or more 5 compounds in the target libraries are identified. Preferably, bound pairs may be identified by selection (e.g., by tagging one of the components, combined with PCR to identify the other). In certain particularly preferred embodiments of this aspect of the invention, the target library or libraries comprise polypeptides and/or proteins. 10295] As described herein, the present invention also provides new modes of nucleic 10 acid-templated synthesis, including simultaneous incompatible reactions and one p'ot multi-step ordered synthesis (e.g., incubating three DNA-linked amino acids and one template so that only a single tripeptide, of specified sequence, is produced). The invention also provides nucleic acid templated synthesis in organic solvents (e.g., methylene chloride, dimethylformamide). [0296] Yet another application of the inventive system is to identify and/or eVolve new 15 templates for nucleic acid-templated synthesis. For instance, the present invention allows identification of nucleic acid templates that, when contacted with reagents that are sufficient to participate in a reaction to generate a selectable product, most efficiently lead to production of that product. [0297] The invention also provides information useful to inform the development of 20 chemical reaction pathways. For instance, according to the present invention, a researcher can select from within a library of nucleic acid-templated substrates those that permit a complex chemical reaction to take place (e.g., macrocyclization, which can be selected for by, for example, loss of a biotin leaving group). When successful reaction conditions have been identified, the inventive system allows ready identification of participating components. Thus, 25 new chemistries can be developed without prior knowledge of the reagents and/or pathways likely to be useful in the reaction. VI. KITS [0298] The present invention also provides kits and compositions for use in the inventive methods. The kits may contain any item or composition useful in practicing the present 30 invention. The kits may include, but are not limited to, templates, (e.g., end-of-helix, hairpin, omega, and T architectures), anticodons, transfer units, monomer units, building blocks, WO 2004/016767 PCT/US2003/025984 -100' reactants, small molecule scaffolds, buffers; solvents, enzymes (e.g., heat stable polymerase, reverse transcriptase, ligase, restriction endonuclease, ekonuclease, Klenow fragment, polymerase, alkaline'phosphatase, poIynuclotide kinase), linkers, protecting groups, polynucleotides, nucleosides, nucleotides, salts,'acids, bases, solid supports, or any combinations 5 thereof. 102991 A kit for preparing unnatural polynr&s should contain items needed to prepare, unnatural polymers using the methods described herein.: Such a kit may include templates,. anti-, codons, transfer units, ionomers units, or combinations thereof A kit foi- synthesizing small molecules may include templates, anti-codons, transfer units, building blocks, small molecule 10 scaffolds, or combinations thereof [0300] The inventive kit can also.be pquipped with items needed to amplify and/or evolve a polynucleotide template such as a heat stable polymerase for PCR, nucleotides, buffer, and primers. In- certain.other embodiments, the inventive kit includes items commonly used in performing DNA shuffling such as polynucleotides, ligase. and nucleotides. 15 10301] In addition to the templates and transfer units described herein, the present invention also includes compositions comprising complex small molecules, scaffolds, or unnatural polymer prepared by any one or more of the methods of the invention as described herein. [03021 A kit for identifying new chemical reactions or functionality may include 20 template associated with reactive units (reactants), transfer units associated with reactive units (reactants), reagents, acids, bases, catalysts, solvents, biotin, avidin, avidin beads, etc. The kit can also include reagents for generating the template associated with a reactive group (e.g., biotin, polynucleotides, reactive units, Klenow fragment of DNA pol I, nucleotides, avidin beads, etc.). The kit can also include reagents for PCR (e.g., buffers, heat stable polymerase, 25 nucleotides, primers, etc.). 10303] The following examples contain important additional information, exemplification and guidance that can be adapted to the practice of this invention in its various embodiments and equivalents thereof.

WO 2004/016767 PCT/US2003/025984 - -101 EXA19P4ES [0304] Examples 1 and 2 describe the preparatiOn of materials for use'in nucdic acid templated synthesis and describe'specific synthetic reactions. Example 3 discusses multi-step synthesis. Example 4 describes the compatibility.of micleic acid-templated synthesis with 5 organic solvents. Example 5 describes specific template architectures useful in the practice of certain DNA-templated syntheses: Example 6 described stereoselectivity in nucl6ic acid. template synthesis.. Example 7 describes the nse of DNA-templated synthesis to direct otherwise incompatible reactions in a single solution. Example 8 describes functional group transformation reactions that can be carried out by nucleic acid-templated synthesis. Example 9 10 describes the synthesis of exemplary compounds and libraries. Example 10 describes the use of polymerases to translate DNA into nonnatural polymers. Example 11 describes in vitro selection protocols. Example 12 describes the application of DNA-ternplated synthesis toward the discovery of new chemical reactions. Example 1: The Generality of DNA-Templated Synthesis 15 [0305] Nucleic acid-templated synthesis is extremely versatile and permits the synthesis of a variety of chemical compounds. This Example demonstrates that it is possible to perform DNA-templated synthesis using two different DNA template architectures. [0306] As shown in Figure 15, templates with a hairpin (H) or end-of-helix (E) architecture bearing electrophilic maleimide groups were prepared to test their reactivity with a 20 transfer unit comprising, a complementary DNA oligonucleotide associated with a thiol reagent. Both the H and E templates reacted efficiently with one equivalent of the DNA-linked thiol reagent to yield the thioether product in minutes at 25 *C. DNA-templated reaction rates (kpp ~1 5 M-'s-1) were similar for H and E architectures despite significant differences in the relative orientation of their reactive groups. In contrast, no product was observed when using reagents 25 containing sequence mismatches, or when using templates pre-quenched with excess P mercaptoethanol (see Figure 15). Thus, both DNA templates support a sequence-specific DNA templated reaction even though the structures of the resulting products differ markedly from the structure of the natural DNA backbone. Little or no non-templated intermolecular reaction products were observed under the reaction conditions (pH 7.5, 25 0 C, 250 mM NaCl, 60 nM 30 template transfer unit), demonstrating the specificity of the DNA-templated reaction.

WV LW4/UIb/6'/ PCT/US2003/025984 -102 [03071 Indeed, sequence-specific DNA-templated reactions spanning a variety of reaction types (SN 2 substitutions, additiodis to a,p-unsaturated carbonyl systems, and additions to vinyl sulfones), nucleophiles thiolss and miness, and reactant structures all proceeded with good yields and excellent sequence selectivity (see, Figure 16)., Matched (M) or mismatched (X) 5 reagents linked to thiols (S) or primary amines (N) were mixed with I equivalent of template functionalized with the variety of electrophiles shown in Figure 16. Reactions with thiol reagents were conducted at pH 7.5 under the following conditions: SLAB and SBAP: 37"C, 16 hours; SIA: 25*C, 16 hours, SMCC, GMBS, BMPS, SVSB: 25 0 C, 10 minutes. Reactions with amine reagents were conducted at 25*C, pH 8.5 for 75 minutes. Expected product masses were 10 verified by mass spectrometry. In each case, matched but not mismatched reagents afforded product efficiently despite considerable variations in'their transition state geometry, steric hindrance, and conformational flexibility. Collectively these findings indicate that nucleic acid templated synthesis is a general phenomenon capable of supporting a range of reaction types, and is not limited to the creation of structures resembling nucleic acid backbones. 15 [0308] Sequence discrimination is important for the faithful translation of a nucleic acid into a synthetic reaction product. To test the sequence discrimination of DNA-templated synthesis, hairpin templates linked to an iodoacetamide group were reacted to thiol-bearing transfer units containing 0, 1, or 3 mismatches. At 25"C, the initial rate of reaction of the thiol bearing transfer unit with no mismatches was 200-fold faster than that of transfer units bearing a O single mismatch (kpp = 2.4 x10 4 N1s- vs. 1.1 x 102 Ms-; Figure 17A). [0309] In addition, small amounts of products arising from the annealing of mismatched reagents could be eliminated by elevating the reaction temperature beyond the melting temperature T. of the mismatched reagents (Figure 17B). In Figure 17B, the reactions in Figure 17B were repeated at the indicated temperatures for 16 hours. The calculated reagent Tm Z5 values were found to be 384C (matched) and 28 0 C (single mismatch). The inverse relationship between product formation and temperature indicates that product formation proceeds by a DNA-templated mechanism rather than by a simple intermolecular mechanism. [03101 In addition to reaction generality and sequence specificity, DNA-templated synthesis, under certain circumstances, also demonstrates remarkable distance independence. 0 Both H1 and E templates linked to maleimide or ct-iodoacetamide groups promoted sequence specific reaction with matched, but not mismatched, thiol reagents annealed anywhere on the WO 2004/016767 PCT/US2003/025984 -103 templates examined thus far (up to 30 bases away from the reactive group on the template). Reactants aninealed one base away reacted viith similar fates as those annealed 2; 3, 4,,6; 8, 10, 15, 20, or 30 bases akvay (Figure, 18). 'The reaction illustrated in Figure 18 used a 41.-base E template and a 10-base reagent designed to anne'al 1-30 bases from the 5'.end of the template. 5 The kinetid profiles of Figure 18 show the average of two trials (deviations <10%). TIe "n = 1 mis" reagent contained three mismatches. In till cases, templated reaction rates were several' hundred-fold higher than the rate of untemplaten (misinatched) reaction (kapp = I0 -10' M's' vs. I~th. the ratect ofr effiteenly forme 5 x 10 Mis4). At intervening distances of 30'bases, products were efficiently formed presunbly through transition states resemblingt400-membered rings. 10 [03111 In order to'further characterize the basis of the distance independence of DNA templated synthesis, a series of modified E terplates wete first synthesized in which the interveningbases were replaced by a series of DNA analogs designed to evaluate the possible contribution of'(i) interbase interactions, (ii) conformational preferences of the DNA backbone, (iii) the charged phosphate backbone, and (iv) backbone hydrophilicity. Templates in-which the 15 intervening bases were replaced with any of the analogs in Figure 19 showed little effect on the rates of product formation. [0312] In the experiment shown in Figure 19, the n = 10 reaction in Figure 18 was' repeated using templates in which the nine bases following the 5'-NH 2 -dT were replaced with the backbone analogues shown. Five equivalents of a DNA oligonucleotide complementary to 20 the intervening bases were added to the "DNA + clamp" reaction. Reagents were either completely matched (0) or contained three mismatches (3). The gel shows reactions after 25 minutes at 25 0 C. Figure 19 shows that the backbone structural elements specific to DNA are not responsible for the observed distance independence of DNA-templated synthesis. However, the addition of a 10-base DNA oligonucleotide "clamp" complementary to the single-stranded 25 intervening region significantly reduced product formation (Figure 19), suggesting that the flexibility of this region is critical to efficient DNA-templated synthesis. [03131 The distance independent reaction rates may be explained if the bond-forming events in a DNA-templated format are sufficiently accelerated relative to their nontemplated counterparts such that DNA annealing, rather than bond formation, is rate-determining. If DNA 30 annealing is at least partially rate limiting, then the rate of product formation should decrease as the concentration of reagents is lowered because annealing, unlike templated bond formation, is WU 2004/116767 PCT/US2003/025984 -104 a bimblecular process. Figure 20 shows the results of experiments in which the n , n = 10, anil n 1 mismatched (mis) reactions described in Figure 18 Were repeated with template and reagent concentrations of 12.5, 25, 62.5 or 125 nM. Figure 20 shows that decreasing thd concentration of reactants in the case of the E template with one or ten intervening bases 5 between reactive groups resulted in' a marked decrease in the observed reaction rate. This observation suggests that proximity effects in DNA-templated synthesis can enhance bond formation rates to the point that DNA annealing becomes rate-determining. {03141 These findings raise the possibility of using DNA-templated synthesis to translate in one pot libraries of DNA into solution-phase libraries of synthetic molecules suitable for PCR 3 amplification and selection. The sequence specificity described above suggests that'mixtures of reagents may be able to react predictably with complementary mixtures of templates. Finally, the observed distance independence suggests that different template codons can be used to encode different reactions without impairing reactions rates. [03151 As a demonstration of this approach, a library of 1,025 maleimide-link6d templates was synthesized, each with a different DNA sequence in an eight-base encoding region (Figures 21A-21B). One of these sequences, 5'-TGACGGGT-3', was arbitrarily chosen to code for the attachment of a biotin group to the template. A library of thiol reagents linked to 1,025 different oligonucleotides was also generated. The reagent linked to 3'-ACTGCCCA-5' , contained a biotin group, while the other 1,024 reagents (transfer units) contained no biotin. Equimolar ratios of all 1,025 templates and 1,025 reagents were mixed in one pot for 10 minutes at 25 0 C and the resulting products were selected in vitro for binding to streptavidin. Molecules surviving the selection were amplified by PCR and analyzed by restriction digestion and DNA sequencing. [0316] Digestion with the restriction endonuclease Tsp45I, which cleaves GTGAC and 5 therefore cuts the biotin encoding template but none of the other templates, revealed a 1:1 ratio of biotin encoding to non-biotin encoding templates following selection. In the experiments shown in Figure 22A, lanes I and 5 represent the PCR-amplified library before streptavidin binding selection; lanes 2 and 6 represent the PCR-amplified library after selection; lanes 3 and 7 represent the PCR amplified authentic biotin-encoding template; and lane 4 represents a 20 bp ) ladder. Lanes 5-7 were digested with Tsp45L DNA sequencing traces of the amplified templates before and after selection are also shown, together with the sequences of the non- WtO 2004/016767 PCT/US2003/025984 -105 biotin-eicoding and biotin-encoding templates. The results summarized in Figure 22A represent a 1;000-fold ,enrichment compared with'the unselected library. DNA sequencing of the PCR amplified.pool before and after selection suggested a similar degree df enrichment and indicated that the biotin-encoding template is the major product after selection and amplification 5 (Figure 22A). The ability of DNA-templated-synthesis to support the simultaneous sequence specific reaction of 1,025 reagents, each of which faces'a 1 024:1 ratio of norn-partner to,partner templates, demonstrates its potential as a ieth4 to create synthetic libraries in one pot. [0317] Taken together, these results show that it, is possible to translate, select, and amplify a synthetic library member having a specific property (for example, bind aVidin) as 10 shown in Figure 22B. Furthermore, these results indicate that nucleic acid-templated synthesis is a surprisingly general phenomenon capable of directing, rather than simply encoding, a range of chemical, reactions to form products unrelated in structure to nucleic: acid backbones. For several reactioits examined, the DNA-ternplated format accelerates the rate 6f bond formation beyond the rate of a 10-base DNA oligonucleotide annealing to its complement, resulting in 15 surprising distance independence. The facile nature of long-distance DNA-templated reactions may also arise in part from the tendency of water to contract the volume of nonpolar reactants (see, C.-J. Li et al. Organic Reactions in Aqueous Media, Wiley and Sons: New York, 1997) and from possible compactness of the intervening single-stranded DNA between reactive groups. Materials and Methods 20 103181 . DNA Synthesis. DNA oligonucleotides were synthesized on a PerSeptive Biosystems Expedite 8909 DNA synthesizer using standard protocols and purified by reverse phase HPLC. Oligonucleotides were quantitated spectrophotonetrically and by denaturing polyacrylamide gel electrophoresis (PAGE) followed by staining with ethidium bromide or SYBR Green (Molecular Probes) and quantitation using a Stratagene Eagle Eye II densitometer. 25 Phosphoramidites enabling the synthesis of 5'-NH 2 -dT, 5' tetrachlorofluorescein, abasic backbone spacer, C3 backbone spacer, 9-bond polyethylene glycol spacer, 12-bond saturated hydrocarbon spacer, and 5' biotin groups were purchased from Glen Research, Sterling, Virginia, USA. Thiol-linked oligonucleotide reagents were synthesized on C3 disulfide controlled pore glass from Glen Research, Sterling, Virginia, USA. 30 [0319] Template Functionalization. Templates bearing 5'-NH 2 -dT groups were transformed into a variety of electrophilic functional groups by reaction with the appropriate WO 2004/016767 PCT/U1S2003/025984 -106 electrophile-N-hydroxysuccinimide (NHS) ester (Pierce, Rockford, IL, USA). Reaptions were performed in 200 mM sodium phosphate'pH 7.2 with 2 mg/mL electrophile-NIS ester, 10% dimethylsulfoxide (DMSO), and up to -106 pLg of 5'-amino template at 25 'C for I hours.' Desired products were purified by reverse-phase HIPLC and characterized by gel electrophoresis and 5 MALDI mass spectrometry. [0320] DNA-templated synthesis' reactions. Reactions were initiated by mixing equimolar quantities of reagent (transfer unit) and template in buffer containing 50 mM N-[3 ,rprpholipopropane]sulfonic acid (MOPS) pH 7.5 aud' 50 mM NaCl at the desired temperature (25 4C unless stated otherwise). Concentrations of reagents and templates were 60 nM unless 10 otherwise indicated. At various time points, aliquots were removed, quenched with excess p mercaptoethanol, and analyzed by denaturing PAGE. Reaction products were quantitated by densitometry using their intrinsic fluorescence or by staining followed by densitometry. Representative products were also verified by MALDI mass spectrometry. [03211 In Vitro Selection for Avidin Binding. Products of the library translation 15 reaction (Figure 21A-21B) were isolated by ethanol precipitation and dissolved in binding buffer (10 mM Tris pH 8, 1 M NaCI, 10 mM ethylenediaminetetraacetic acid (EDTA)). Products were incubated with 30 tg of streptavidin-linked magnetic beads (Roche Biosciences) for 10 minute at room temperature in 100 pL total volume. The beads were washed 16 times with binding buffer and eluted by treatment with I pmol free biotin in 100 uL binding buffer at 20 70 'C for 10 minutes. The eluted molecules were isolated by ethanol precipitation and amplified by standard PCR protocols (2 mM MgC12, 55 'C annealing, 20 cycles) using the primers 5' TGGTGCGGAGCCGCCG [SEQ ID NO: 35] and 5' CCACTGTCCGTGGCGCGACCCCGGCTCC TCGGCTCGG [SEQ ID NO: 36]. Automated DNA sequencing used the primer 5-CCACTGTCCGTGGCGCGACCC [SEQ ID NO: 371. 25 [03221 DNA Sequences. Sequences not provided in the Figures are as follows: matched reagent in Figure 16 SIAB and SBAP reactions: 5'-CCCGAGTCGAAGTCGTACC-SH [SEQ ID NO: 38]; mismatched reagent in Figure 16 SIAB and SBAP reactions: 5' GGGCTCAGCTTCCCCATAA-SH [SEQ ID NO: 39]; mismatched reagents for other reactions in Figures 16, and 17A-17B; 5'-FAAATCTTCCC-SH (F= tetrachlorofluoreseein) [SEQ ID 30 NO: 40]; reagents in Figure 16 containing one mismatch: 5'-FAATTCTTACC-SH [SEQ ID NO: 41]; E templates in Figures 15 and 16 SMCC, GMBS, BMPS, and SVSB reactions, and WO 20041016767 PCT/UfS2003/025984 -107 Figures 17A-17B: 5(NHdT) CGCGAGCGTACGCTCGCGATGGTACGAATTCGACTCGGGAATAC CACCTTCGACTCGAGG [SEQ ID NO: 42]; H template in Figure 16 SLAB, SBAP, and SIA reactions: 5-(NH2dT)- CGCGAGCGTACGCTCGCGATGGTACGAATTC [SEQ ID NO: 43]; 5 clamp oligonucleotide in Figure 19: 5-ATTCGTACCA [SEQ ID NO: 44] Example 2: Exemplary Reactions for Use in DNA-Templated Synthesis [0323] This Example demonstrates that DNA-templated synthesis can direct a modest collectiont of chemical reactions without requiring the precise alignment of reactive groups into DNA-like conformations. Furthermore, this Example also demonstrates that it is possible to 10 simultaneously translate in one-pot a library of more than 1,000 templates into the corresponding thioether products, one of which could be enriched by in vitro selection for binding to streptavidin and amplification by PCR. [0324] As described in detail herein, a variety of chemical reactions for example, DNA templated organometallic couplings and carbon-carbon bond forming reactions other than 15 pyrimidine photodimerization can be utilized to construct small molecules. These reactions represent an important step towards the in vitro evolution of non-natural synthetic molecules by permitting the DNA-templated construction of a diverse set of structures. [0325} The ability of DNA-templated synthesis to direct reactions that require a non DNA-linked activator, catalyst or other reagent in addition to the principal reactants has also 20 been demonstrated herein. To test the ability of DNA-templated synthesis to mediate such reactions without requiring structural mimicry of the DNA-templated backbone, DNA-templated reductive aminations between an amine-linked template (1) and benzaldehyde- or glyoxal-linked reagents (3) with millimolar concentrations of sodium cyanoborohydride (NaBH3CN) at room temperature in aqueous solutions can be performed (see, Figure 23A). Significantly, products 25 formed efficiently when the template and reagent sequences were complementary, while control reactions in which the sequence of the reagent did not complement that of the template, or in which NaBH 3 CN was omitted, yielded no significant product (see Figures 23A-23D and 24). Although DNA-templated reductive animations to generate products closely mimicking the structure of double-stranded DNA have been previously reported (see, for example, Li et aL 30 (2002) J. AM. CHEM. Soc. 124: 746 and Gat et ai (1998) BIOPoLYMERS 48: 19), these results WO 2004/016767 PCTIUS2003/025984 -108.

demonstrate that reductive amination to generate struct-ures unrelated to the phosphoribose backbone can take place efficiently and, squence-specifically. [0326} Referring to Figures 25A-25B, DNA-templated armide bond formations between amine-linked templates 4 and 5 and carboxylate-linked reagents 6-9 mediated by 1-(3 5 dimethylaminopropyl) 3-ethylcarbodiinnide (PDC) and N-hydroxylsulfosuccinimide (sulfo NHS) generated aide products in good yields at pH 6,0, 25 0 C. Product fornatiori was (i) sequence-specific, (ii)-dependent on thepresence of F1Db, and (iii) insensitive to the steric encumbrance of the amine or carboxylate. Efficient DNA-templated amide formation Was also mediated by the water-stable activator 4-(4,6-dimethoxy-1,3,5-trizin-2-yl)-4 10 methylmorpholinium chloride (DMT-MM) instead of EDC and sulfo-NHS (Figures 24 and 25A-25B)- The efficiency and generality of DNA-templated amide bond formation under these conditions, together with the large number of commecially available chiral amines and carboxylic acids, make this reaction an attractive candidate in future DNA-templated syntheses of structurally diverse small molecule libraries. 15 [0327j Carbon-carbon bond forming reactions are also important in both chemical and biological syntheses and thus several such reactions can be utilized in a nucleic acid-templated format. Both the reaction of nitroalkane-linked reagent (10) with aldehyde-linked template (11) (nitro-aldol or Henry reaction) and the conjugate addition of 10 to maleimide-linked template (12) (nitro-Michael addition) proceeded efficiently and with high sequence specificity at pH 7.5 20 8.5, 25 0 C (Figures 23A and 24). In addition, the sequence-specific DNA-templated Wittig reaction between stabilized phosphorus ylide reagent 13 and aldehyde-linked templates 14 or 11 provided the corresponding olefin products in excellent yields at pH 6.0-8.0, 25 0 C (Figures 23B and 24). Similarly, the DNA templated 1,3-dipolar cycloaddition between nitrone-linked reagents 15 and 16 and olefin-linked templates 12, 17 or 18 also afforded products sequence 25 specifically at pH 7.5, 25"C (Figures 23B, 23C and 24). [0328] In addition to the reactions described above, organometallic coupling reactions can also be utilized in the present invention. For example, DNA-templated Heck reactions were performed in the presence of water-soluble Pd precatalysts. In the presence of 170 mM Na 2 PdCb, aryl iodide-linked reagent 19 and a variety of olefin-linked templates including 30 maleimide 12, acrylamide 17, vinyl sulfone 18 or cinnamamide 20 yielded Heck coupling products in modest yields at pH 5.0, 25 0 C (Figures 23D and 24). For couplings with olefins 17, WU 20t1016767 PCT/US20031025984 109 18 and 20, adding two equivalents of P(p-SO 3 C6H4) 3 per equivalent of Pd prior to template and reagent addition typically increased overall yields by 2-fold Control reactions containing sequence mismatches or lacking Pd predatalyst yielded no 'product. [03291 Example 1 above shows that certain DNAL-templated reactions demonstrate 5 distance independence. Distance independence may arise when the rate of bond formation in the DNA-templated reaction is greater than tlie'rate of timplate-reagent annealing. Although only a subset of chemistries fall into this category, any DNA-templated reaction that affords' corhparable product yields when the reagent is annbaled at various distances from the reactive end of the template is of special interest because it can be encoded at a variety of template 10 positions. In order to evaluate the ability of the DNA-templated reactions developed in this Example to take place efficiently when reactants are separated by distances relevant to library encoding, the yields of reductive amination, amide formation, nitro-aldol addition, nitro-Michael addition, Wittig olefination, dipolar cyckoaddition, and Heck coupling reactions were compared when either zero (n = 0) or ten (n = 10) bases separated the annealed reactive groups,(Figure 15 26A). Among the reactions described here or in Example 1, aide bond formation, nitro-aldol addition, Wittig olefination, Heck coupling, conjugate addition of thiols to maleimides and SN 2 reaction between thiols and a-iodo amides demonstrate comparable product formation when reactive groups are separated by zero or ten bases (Figure 26B). Figure 26B shows the results of denaturing polyacrylamide gel electrophoresis of a DNA-templated Wittig olefination 20 between complementary 11 and 13 with either zero bases (lanes 1-3) or ten bases (lanes 4-6) separating the annealed reactants. Although the apparent second order rate constants for the n = 0 and n = 10 reactions differ by three-fold (kapp (n = 0) = 9.9 x 103 m- 1 s while kapp (n = 10) = 3.5 x 103 M-'s), product yields after 13 hours at both distances were nearly quantitative. Control reactions containing sequence mismatches yielded no detectable product. These 25 findings indicate that these reactions can be encoded during synthesis by nucleotides that are distal from the reactive end of the template without significantly impairing product formation. [0330] In addition to the DNA-templated SN2 reaction, conjugate addition, vinyl sulfone addition, amide bond formation, reductive amination, nitro-aldol (Henry reaction), nitro Michael, Wittig olefination, 1,3-dipolar cycloaddition and Heck coupling reactions described directly 30 above, a variety of additional reagents can also be utilized in the method of the present invention. For example, as depicted in Figure 27, powerful aqueous DNA-templated synthetic reactions oi / PCT/US2003/025984 including, but notlimited to, the Lewis acid-catalyged aldol addition, Mannich reaction, Robinson annulation reactions, additions of allyl ihdiunh-zinc and tin to ketones and aldehydes; Pd-assisted allylic substitution, DieIs.Aldercycloadditions, and'hetero-Di6i-Alder reactions can be utitlzed efficiently in aqueous solvent and ar'e important complexity-building reactions. 5 [03311 Taken together, these results expand considerably the reaction scope of DNA templated synthesis. A wide variety of i-eactions can proced efficiently and selectively when, the corresponding reactants are programmed with complementary sequences. By augmenting the repertoire of khown DNA-templated reactions to include carbon-carbbn bond forming and organoretallic reactions (nitro-aldol additions; nitro-Michael additions, Wittig olefinations 10 dipolar cycloadditions, and Heck couplings) in addition to previously reported amide bond formation (see, Schmidt et al. (1997) NUcttic AciDs Ras. 25: 4792; Bruick et al (1996) CHEM. BIOL. 3: 49), imine formation (Czlapinski'et aL (2001 J. AM. CHEM. SOC. 123: 8618), reductive aminatiorn (Li et al (2002) J. AM. CHEM. SOc. 124; 746; Gat et aL (1998) BIOPOLYMERS 48: 19),

SN

2 reactions (Gartner et aL (2001) J. AM. OSPEM. SOC. 123: 6961; Xu et al. (2001) NAT., l5 BIOTECHNOL. [9: 148; Herrlein et al. (1995) J. AM. CHEM. Soc. 117: 10151) conjugate addition of thiols (Gartner et al. (2001) J. AM. CHEM. SOC. 123: 6961), and phosphoester or phosphonamide formation (Orgel et al. (1995) Acc. CHEM. RES. 28: 109; Luther et al (1998) NATURE 396: 245), these results may permit thesequence-specific translation of libraries of DNA into libraries of structurally and functionally diverse synthetic products. .0 103321 Because minute quantities of templates encoding desired molecules can be amplified by PCR, the yields of DNA-templated reactions arguably are less critical than the yields of traditional synthetic transformations. Nevertheless, many of the reactions discussed in this Example proceed efficiently. Materials and Methods 25 [0333] Functionalized templates and reagents were typically prepared by reacting 5'-NH2 terminated oligonucleotides (for template 1), 5'-NH 2

-(CH

2 0) 2 terminated oligonucleotides (for all other templates) or 3'-OPO-CH 2 CH(CH2oH)(CH 2

)

4

NH

2 terminated nucleotides (for all reagents) with the appropriate NHS esters (0.1 volumes of a 20 mg/mL solution in DMF) in 0.2 M sodium phosphate buffer, pH 7.2, 25 0 C, for I hour to provide the template and reagent 0 structures shown in Figures 23A-23D and 25A-25B. For amino acid linked reagents 6-9, 3' OPO3CH 2

CH(CH

2 OH)(CH1 2

)

4

NH

2 terminated oligonucleotides in 0.2 M sodium phosphate WO 2004/016767 PCT/U1S2003/025984 - 111 buffer, pH 7.2 were reacted with 0.1 volumes of a 1.00 mM bis[2 (succiniiidyloxycarbonyloxy)ethyl]sulfone (BSOCOES, Pierce, Rockford, IL, USA) solution in DMF for 10 minutes at 25 0 C, followed by 0.3 volumes of a 300 mM amino acid in 300 mM sodium hydroxide (NaOH) for 30 minutes at 25 0 C. 5 [0334] Functionalized templates and reagents were purified by gel filtration using Sephadex G-25 followed by reverse-phas 6 IIPLC (0.1 triethylainjmonium acetate-acetonitrile gradient) and characterized by MALDI mass spectrometry. {03351 For the DNA templated reactions described in Figures 23A-23D; reactions were conducted at 25CC with one equivalent each of template and reagent at 60 nM final concentration 10 unless otherwise specified. Conditions: (a) 3 mM NaBH 3 CN, 0.1 M N-[2-morpholinoethane] sulfonic acid (MES) buffer pH 6.0, 0.5 M NaCl, 1.5 hours; b) 0.1 M N-tris[hydroxymethylj methyl-3-aminopropanesulfonic acid (TAPS) buffer pH 8.5, 300 mM NaCl, 12 hours; c) 0.1 M pH 8.0 TAPS buffer, 1 M NaCl, 5 0 C, 1.5 hours; d) 50 mM MOPS buffer pH 7.5, 2.8 M NaCl, 22 hours; e) 120 nM 19, 1.4 mM Na 2 PdC 4 , 0.5 M NaOAc buffer pH 5.0, 18 hours; (f) Premix 15 Na 2 PdCl4 with two equivalents of P(p-S0 3

C

6

H

4

)

3 in water for 15 minutes, then add to reactants in 0.5 M NaOAc buffer pH 5.0, 75 mM NaCl, 2 hours (final [Pd] 0.3 mM, [19] = 120 nM). The oldfin geometry of products from 13 and the regiochemistries of cycloaddition products from 14 and 16 are presumed but not verified (Figures 23A-23D). Products were characterized by denaturing polyacrylamide gel electrophoresis and MALDI mass spectrometry. For all 20 reactions under the specified conditions, product yields of reactions with matched template and reagent sequences were greater than 20-fold higher than that of control reactions with scrambled reagent sequences. 103361 The conditions for the reactions described in Figures 25A-25B were: 60 nM template, 120 nM reagent, 50 mM DMT-MM in 0.1 M MOPS buffer pH 7.0, 1 M NaCl, for 16 25 hours at, 25 0 C; or 60 nM template, 120 nM reagent, 20 mM EDC, 15 nM sulfo-NHS, 0.1 M MES buffer pH 6.0, 1 M NaCl, for 16 hours at 25 0 C. In each row of the table in Figures 25A 25B, yields of DMT-MM-mediated reactions between reagents and templates complementary in sequence were followed by yields of EDC and sulfo-NHS-mediated reactions. In all cases, control reactions with mismatched reagent sequences yielded little or no detectable product and 30 products were characterized by denaturing polyacrylamide gel electrophoresis and MALDI mass spectrometry.

WO) 2004/016767 PCT/U S20031025984 .- 112 [03371 Figure 24 depicts the analysis by denaturing polyacrylamide gel electrophoresis of representative DNA-templated reactions-lised jn Figures 23A-23D and 25A-25B. The structures of reagents and templates correspbnd to the numbering in Figuts 23A-23) and 25A 25B. Lanes 1, 3, 5, 7, 9, 11: reaction of matched (complementary or "M") reagents and 5 templates under conditions listed in Figures 23A23D and 25A-25B (the reaction between 4 and 6 was mediatedby DMT-MM). Lanes 2, 4, 6, 8, 10, 12: reaction of mismatched (non complementary or "X"3 reagents and templatesInder'conditions identical to those in lanes 11 3, 5, 7, 9 aid 11, respectiVely. [0338- The sequences of oligonucleotide templates and reagents are as follows (5' to 3' 10 direction, n-refers to the lumber of bases between reactive groups when template and reagent are annealed as shown in Figure 26A). 1: TGGTACGAATTCGACTCGGG [SEQ ID NO: 45]; 2 and 3 matched: GAGTCGAATTCGTACC [SEQ ID NO: 46); 2 and 3 mismatched: GGGCTCAGCTTCCCCA [SEQ ID NO: 47]; 4 and 5: GGTACGAATTCGACTCGGGAATACCACCTT [SEQ ID NO: 48]; 6-9 matched (n = 10): 15 TCCCGAGTCG [SEQ ID NO: 49]; 6 matched (n = 0): AATTCGTACC [SEQ ID NO: 501; 6-9 mismatched: TCACCTAGCA [SEQ JID NO: 51]; 11, 12, 14, 17,18, 20: GGTACGAATTCGACTCGGGA [SEQ ID NO: 52]; 10, 13,16,19 matched: TCCCGAGTCGAATTCGTACC [SEQ ID NO: 531; 10,13,16,19 mismatched: GGGCTCAGCTTCCCCATAAT [SEQ ID NO: 541; 15 matched: AATTCGTACC [SEQ ID 20 NO: 55]; 15 mismatched: TCGTATTCCA [SEQ ID NO: 56]; template for n = 10 vs. n = 0 comparison: TAGCGATTACGGTACGAATTCGACTCGGGA [SEQ ID NO: 57]. {0339] Reaction yields were quantitated by denaturing PAGE followed by ethidium bromide staining, UV visualization, and charge-coupled device (CCD)-based densitometry of product and template starting material bands. Yield calculations assumed that templates and 25 products stained with equal intensity per base; for those cases in which products were partially double-stranded during quantitation, changes in staining intensity may have resulted in higher apparent yields. Example 3: Multi-Step Small Molecule Synthesis Programmed by DNA Templates [0340] This Example demonstrates that it is possible to perform multi-step small 30 molecule synthesis via DNA-templated chemistries.

WO 2004/016767 PCT/US2003/025984 -' - 113 [0341] DNA-templated synthesis can direct a wide variety of poweiful chemical reactions with high sequence-specificity-and without requiring structural mimicry bf the DNA backbone. The application of this approach to synthetic molecules of useful complexity, however, requires the development of general methods to permit the product of a DNA 5 templated reaction to'uridergo subsequent DNA-ternplated transformations. [0342] ' ulti-stepDNA-templated siall molecue synthesis faces tw6 major challenges, beyond those associated with DNA-templated synthesis in general. First, the DNA used to direct reagents td appropriate templates must be removed fpom the product of a DNA-templated reaction prior to subsequent DNA-templated synthetic step#j in order to prevent und6sired 10 hybridization to the template. Second, multi-step synthesis often requires the purification and isolation of intermediate products. To address these challenges, three distinct strategies have been developed (i) to link chemical reagents (reactive units) with their decoding DNA oligonucleotid6s and (ii) to purify product after any DNA-templated synthetic step. [03431 When possible, an ideal reagent-oligonucleotide linker for DNA-templated IS synthesis positions the oligonucleotide as a leaving group of the reagent. Under this 'autocleaving" linker strategy, the oligonucleotide-reagent bond is cleaved as a natural chemical consequence of the reaction (see, Figure 28A). [0344] As the first example of this approach applied to DNA-templated chemistry, a dansylated Wittig phosphorane reagent (1) was synthesized in which the decoding DNA 20 oligonucleotide was attached to one of the aryl phosphine groups (Hughes (1996) TETRAHEDRON LETT, 37: 7595). DNA-templated Wittig olefination with aldehyde-linked template 2 resulted in the efficient transfer of the fluorescent dansyl group from the reagent to the template to provide olefin 3 (Figure 28A). As a second example of an autocleaving linker, DNA-linked thioester 4, when activated with Ag(f) at pH 7.0 (Zhang et al. (1999) J. AM. CHEM. Soc. 121: 3311) acylated 25 amino-terminated template 5 to afford amide product 6 (Figure 28B). 10345] Ribosomal protein biosynthesis uses ainnoacylated tRNAs in a similar autocleaving linker format to mediate RNA-templated peptide bond formation. To purify desired products away from unreacted reagents and from cleaved oligonucleotides following DNA-templated reactions using autocleaving linkers, biotinylated reagent oligonucleotides and 30 washing crude reactions with streptavidin-linked magnetic beads (see, Figure 30A) were utilized. Although this approach does not separate reacted templates from unreacted templates, WO 2004/016767 PCT/US2003/025984 -114 unreacted templates can be removed in subsequent DNA-templated reaction and purification steps. [0346} Reagents bearing more than one functional group can be linked to their decoding DNA oligonucleotides through second and third linker strategies. In the "scarless linker" 5 approach (Figure 28C), one functional group of the reagent is reserved for DNA-templated bond formation, while the second functional group is used to attach a linker that can be cleaved without introducing additional unwanted chemical functionality. The DNA-templated reaction thpn is fo)lowed by cleavage of the linker attached through the second functional group to afford desired products (Figure 28C). For example, a series of aminoacylation reagents such as (D) 10 Phe derivative 7 were synthesized in which the a-amine is connected through a carbarnoylethylsulfone linker (Zarling et aL (1980) J. IMMUNOLOGY 124: 913) to its decoding DNA oligonucleotide. The product (8) of DNA-templated aide bond formation using this reagent and an amine-terminated template (5) was treated with aqueous base to effect the quantitative elimination and spontaneous decarboxylation of the linker, affording product 9 15 containing the cleanly transferred amino acid group (Figure 28C). This sulfone linker is stable in pH 7.5 or lower buffer at 25 *C for more than 24 hours yet undergoes quantitative cleavage when exposed to pH 11.8 buffer for 2 hours at 37 C. [0347] In some cases it may be advantageous to introduce one or more atoms new, chemical groups as a consequence of linker cleavage. Under a third linker strategy, linker 20 cleavage generates a "useful scar" that can be functionalized in subsequent steps (Figure 28C). As an example of this. class of linker, amino acid reagents such as the (L)-Phe derivative 10 were generated linked through 1,2-diols (Fruchart et al. (1999) TETRAHEDRON LETT. 40: 6225) to their decoding DNA oligonucleotides. Following DNA-templated aide bond formation with amine terminated template (5), this linker was quantitatively cleaved by oxidation with 50 mM aqueous 25 sodium periodate (NaIO 4 ) at pH 5.0 to afford product 12 containing an aldehyde group appropriate for subsequent functionalization (for example, in a DNA-templated Wittig olefination, reductive amination, or nitrolaldol addition). [0348] Figure 29 shows the results of exemplary DNA-templated synthesis experiments using autocleaving linkers, scarless linkers, and useful scar linkers. The depicted reactions were 30 analyzed by denaturing PAGE. Lanes 1-3 were visualized using UV light without DNA staining; lanes 4-10 were visualized by staining with ethidium bromide following by UV WO 2004/016767 PCT/US2003/025984 115' transillumination Conditions forI to 3 were: one equiyalent each of reagent and template, 0.1 M TAPS buffer pH -8.5, 1 M NaClat 25 0 C fox 1 .5 houts. Conditions for 4 to 6 were: three equivalents of-4, 0 1 M MES buffer pH 7.0,"1 M sodium nitrite (NaNO 2 ) 10 mM silver nitrate (AgNO 3 ), at 37 *C for 8 hours. Conditions for 8 to 9 wete 0.1 M 3- (cyclohexylamino)-1 5 propanesulfonic acid (CAIPS) buffer piH 11.8; 60 mM Pmercaptoethanol (BME), at 37 ,C for 2 hours. Finally,'conditions for 11 to 12 were: 50 mM -aqueous NaIO 4 , at 25 *C for 2.honrs. R= NH(CH2) 2 NHfdansyl; Rz.= biotin. 103491 Desired products generated from DNAAt,emplated reactions using the scarless, or useful scar linkers can be readily purified using biotinylated reagent oligonucleotides (Figuret 10 30B). Reagent oligonucleotides together with desired products are first captured on streptavidin linked magnetic beads. Any unreacted temiplate bound to reagent by base pairing is removed by washing the beads with buffer containing 4 M guanidinium chloride. Biotinylated molecules remain bound to the streptavidin beads under these conditions. Desired prokluct then is isolated in pure form. by eluting the beads with linked cleavage buffer (in the examples above; either pH 15 11 or sodium periodate (NaIO 4 )-containing buffer), while reacted and unreacted reagents remain bound to the beads. [03501 As one example of a specific library generated as described above, three iterated cycles of DNA-templated arnide formation, traceless linker cleavage, and purification with streptavidin-linked beads were used to generate a non-natural tripeptide (Figures-31A-B). Each 20 amino acid reagent was linked to a unique biotinylated 10-base DNA oligonucleotide through the sulfone linker described above. The 30-base amine-terminated template programmed to direct the tripeptide synthesis contained three consecutive 10-base regions that were complementary to the three reagents, mimicking the strategy that would be used in a multi-step DNA-templated small molecule library synthesis. 25 [0351] In the first step, two equivalents of 13 were activated by treatment with 20 mM EDC, 15 mM sulfo-NHS, 0.1 M MES buffer pH 5.5, and 1 M NaCl, for 10 minutes at 25 'C. The template then was added in 0.1 M MOPS pH 7.5, and 1M NaC1, at 25 0 C and was allowed to react for 1 hour. The free amine group in 14 then was elaborated in a second and third round of DNA-templated amide formation and linker cleavage to afford dipeptide 15 and tripeptide 16 30 using the following conditions; two equivalents ofreagent, 50 mM DMT-MM, 0.1 M MOPS buffer pH 7.0, 1 M NaCl, at 25 *C for 6 hours. Desired product after each step was purified by WO 2004/016767 PCT/US2003/025984 -116 capture on.avidin-linked beads and elution with 0.1. M CAPS buffer pH 11.8, 60 mM BME, at 37 oCfor 2 hours. The progress of each reaction and purification was followed by denaturing polyacrylamide gel electrophoresis (Figire 31B, bottom)'. Lanes 3, 6, and 9 represent control reactions using reagents containing scrambled oligonucleotide sequences. 5 [0352} The progress of each reaction, purification, and sulfone linker cleavage.step was followed by denaturing polyacrylarnide gel electroplioresis. The .final tripeptide linked to template 16 was digested with the restriction endonublease EcoRI and the digestion fragment 'cntaining the tripeptide was characterized by MAILDImass spectrometry. Beginning with 2 nmol (- 20 pig) of starting material, sufficient tripeptide product was generated to serve as the 10 template for more than 106 in vitro selections and PCR reactions (Kramer et at (1999) CURRENT PROTOCOLS IN MOL. BIOL. 3: 15.1) (assuming 1/10,000 molecules survive selection). No significant product was generated when the starting material template was capped with acetic anhydride, or when control reagents containing sequence mismatches were used instead of the complementary reagents (Figure 31B).

15 [0353] A non-peptidic multi-step DNA-templated small molecule synthesis that uses all three linker strategies developed above was also performed (Figure 32A-32B). An amine terminated 30-base template was subjected to DNA-templated aide bond formation using an aminoacyl donor reagent (17) containing the diol linker and a biotinylated 10-base oligonucleotide to afford amide 18 (two equivalents 17 in 20 mM EDC, 15 mM sulfo-NHS, 0.1 20 M MES buffer pH 5.5, 1 M NaCI, 10 minutes, 25 'C, then add to template in 0.1 M MOPS pH 7.5, 1M NaC1 at 16"C for 8 hours). The desired product then was isolated by capturing the crude reaction on streptavidin beads followed by cleaving the linker with NaO4 to generate aldehyde 19. The DNA-templated Wittig reaction of 19 with the biotinylated autocleaving phosphorane reagent 20 afforded fumaramide 21 (three equivalents 20, 0.1 M TAPS pH 9.0, 3 M NaCl at 25 25 'C for 48 hours). The products from the second DNA-templated reaction were partially purified by washing with streptavidin beads to remove reacted and unreacted reagent. In the third DNA-templated step, furmaramide 21 was subjected to a DNA-templated conjugate addition (Gartner et al. (2001) J. AM. CHEM. SOC. 123: 6961) using thiol reagent 22 linked through the sulfone linker to a biotinylated oligonucleotide (three equivalents 22, 0.1 M TAPS pH 8.5, 1 M 30 NaCl at 25 0 C for 21 hours). The desired conjugate addition product (23) was purified by immobilization with streptavidin beads. Linker cleavage with pH 11 buffer afforded final WV Z4/l// rUT/U 2003/02)984 -117 product 24 in 5-10% overall isolated yield for the three bond forming reactions, two linker cleavage stdps, and three purificatibns (Figures 3A-32B). 10354] The final product was digested with EcoRI and the mass of the small inolecule linked template fragment was confirmed by MALDI mass spectrometry (exact mass: 2568, 5 observed mass: 25661L). As in the tripeptide exafnplejeach of the three reagents used'during this multi-step synthesis, annealed at a unique.location on the DNA template; and control reactions with sequence mismatches yielded no product (Figure 32B, bottom). In Figure 32B,, bottom lanes 3, 6, and 9 represent control reactions, As expected, control reactions in which the Wittig reagent was omitted (step 2) ,lso did not generate product following the third step. 10 [0355] Taken together, the DNA-templated syntheses of compounds 16 and 24 demonstrate the ability of DNA to direct the sequence-programmed multi-step synthesis of both oligomeric and non-oligomeric small mol'eues: unrelated in structure'to nucleic acids. Example 4: Exemplary Reactions in Organic Solvents (03561 As demonstrated herein, a variety of DNA-templated reactions can occur in 15 aqueous media. It has also been discovered that DNA-templated reactions can occur in organic solvents, thus greatly expanding the scope of DNA-templated synthesis. Specifically, DNA templates and reagents have been complexed with long chain tetraalkylammonium cations (see, Jost et al. (1989) NUCLEIC ACIDs REs. 17: 2143; Mel'nikov et a. (1999) LANGMTIR 15: 1923 1928) to permit quantitative dissolution of reaction components in anhydrous organic solvents 20 including CH 2

CI

2 , CHC 3 , DMF and methanol. Surprisingly, it was found that DNA-templated synthesis can indeed occur in anhydrous organic solvents with high sequence selectivity. [0357] Figure 33 shows DNA-templated amide bond formation reactions where the reagents and templates are complexed with dimethyldidodecylammonium cations either in separate vessels or after preannealing in water, lyophilized to dryness, dissolved in CH 2

CI

2 , and 25 mixed together. Matched, but not mismatched, reactions provided products both when reactants were preannealed in aqueous solution and when they were mixed for the first time in CH 2 C1 2 (Figure 33). DNA-templated aide formation and Pd-mediated Heck coupling in anhydrous DMF also proceeded sequence-specifically. [0358] These observations of sequence-specific DNA-templated synthesis in organic 30 solvents imply the presence of at least some secondary structure within tetraalkylammonium- WO 2004/016767 PCT/US2003/025984 -118 complexed DNA in organic media, and should permit DNA receptors and catalysts, to be evolved towards stereoselective binding or catalytic properties in organic solvents. Specifically, DNA templated reactions that are known to occur in aqueous nedia, including conjugate additions, cycloadditions, displacement reactions, and Pd-mediated couplings can also be performed in , 5 organic solvents. [03591 It is contemplated that reactibns in organic solvents may be utilized that are inefficient or impossible to perform in water. For example, while Ru-catalyzed olefin metathesis in water bas been reported (Lynn et at. (1998) J. AM. OHEM. Soc. 120: 1627-1628; Lynn et al. (2000) J. AM. CHEM. Soc. 122: 6601-6609; Mohr e at. (1996) ORGANOMETALLICs 15: 4317 10 4325), the aqueous metathesis system is extremely sensitive to the identities of the functional g-oups. The functional group tolerance of Ru-catalyzed olefin metathesis in organic solvents, however, is significantly more robust. Some exemplary reactions to utilize in organic solvents include, but are not limited to 1,3-dipolar cycloaddition between nitrones and olefins which can proceed through transition states that are less polar than ground state starting materials. 15 Example 5: New Architectures for Nucleic Acid-Templated Synthesis [03601 This Example discloses two different template architectures that further expand the scope of nucleic acid-templated synthesis. [03611 During a nucleic acid-templated chemical reaction a portion of a template anneals to a complementary sequence of an oligonucleotide-linked reagent, holding functional groups on 20 the template and transfer unit in reactive proximity. Template architecture can have a profound effect on the nature of the resulting reaction, raising the possibility of manipulating reaction conditions by rationally designing template-reagent complexes with different secondary structures. [03621 During the course of DNA templated synthesis using the end-of-helix ("E") and 25 hairpin ("H") templates (see, Example 1), two challenges emerged. First, some DNA-templated reactions do not proceed efficiently when the annealed reactive groups on the template and transfer unit (reagent) are separated by even small numbers of bases. Using the E or H architectures, "distance-dependent" reactions can only be encoded by template bases at the reactive end of the template. Second, the presence of double-stranded DNA between annealed 30 reactive groups can greatly reduce the efficiency of templated reactions because, under certain circumstances a single-stranded template may need to be flexible. This may preclude the WO 2004/016767 PCT/US2003/025984 - 119 possibility of performing two or more reactions in a single DNA-templated step using the E or H architectures even though the template oligonucleotide may contain enough bases to encode multiple reactions. This Example discuses two new templite architectures, which -overcome each of these challenges. 10363 ' It was hypothesized that the distance dependence of certain DNA-templated reactions such as 1,3-dipolar cycloadditions and redictiveamirniation could be overcome by designing a new architecture that permits a reagent t6 anneal to two distinct and spatially separated regions of the template. In the "omega" or 'b" architecture (see, Figure 7), the template oligonucleotide contains a small number of constant bases at, for example, the reactive 10 5' end of the template in addition to distal coding regions. The oligonucleotide of the transfer unit for the £ architecture contains at its reactive 3' end the bases that complement the constant region of the template followed by bases that complement a coding region anywhere on the template. The constant regions were designed to be of insufficient length to anneal in the absence of a complementary coding region. When the coding region of the template and transfer 15 unit are complementary and anneal, the elevated effective polarity of the constant regions induces their annealing. Constant region annealing fonrs a bulge (resembling an £) in the otherwise double-stranded template-reagent complex and places groups at the ends of the template and reagent in reactive proximity. This design permits distance-dependent

DNA

templated reactions to be encoded by bases distal from the reactive end of the template. 20 103641 The efficiency of DNA-templated synthesis using the £ architecture was compared with that of the standard E and H architectures. The El architectures studied comprise (i) three to five constant bases at the 5' end of the template followed by (ii) a five- to 17-base loop and (iii) a ten-base coding region. As a basis for comparison, four different classes of DNA-templated reactions were performed that collectively span the range of distance 25 dependence observed to date, [0365] Amine acylation reactions are representative of distance independent reactions that proceed efficiently even when considerable distances (e.g., 30 bases) separate the amine and carboxylate groups. As expected, amine acylation (20 mM DMT-MM, pH 7.0, at 30 "C for 12 hours) proceeded efficiently (46-96% yield) in all architectures with both small and large 30 distances between reactive groups on the reagent and template (Figure 34, lanes 1-5; and Figure 35A). The 0 architecture mediated efficient amine acylation with three, four, or five constant WO 2004/016767 PCT/US2003/025984 -120' bases at the reactive ends of the template and rdagint and 10 or 20 bases between annealed reactants (n = 10 or'20). Importantly, control reactions in which the distal coding region contained three sequence mismatches'failed to generAte significant produo despite the presence of the omplementary three-to five-base constant regions at the ends of the template and reagent 5 (see, Figu're 34, lane 5'for-a representative example). The Q architecture, therefore, did not impede the efficiency or,sequence-sp'ecificity of the distance-independent amine acylafon reaction. [03661 DNA-templated Wittig olefinationreactions proceed at a significantly lower rate when the aldehyde and phosphorane are separated by larger numbers of template bases, even, 10 though product yields.typically are excellent after 12 hpurs or more of reaction regardless of intervening distance. After only 2 hours of reAition.(pH 7.5, 30 -C) in the E or H architectures, however, yields of olefin products were three- to six-fold lower when reactants weie separated by ten or more bases (n 10 or 20) than when reactants are separated by only one base (n = 1) (Figure 34 lanes 6-7, and Figure 35B). In contrast, the 0 architecture with four or five constant 15 bases at the reactive end resulted in efficient and sequence-specific Wittig product formation after 2 hours of reaction even when 10 or 20 bases separated the coding region and reactive end of the template (Figure 34, lanes 8-9, and Figure 35B). These results suggest that the constant regions at the reactive ends of the template and transferunit in the 0 architecture permit the aldehyde and phosphorane moieties to react at an effective concentration comparable to that 20 achieved with the B-architecture when n = 1 (Figure 34). 103671 Among the many DNA-templated reactions studied to date, the 1,3-dipolar cycloaddition and reductive amination reactions demonstrate the most pronounced distance dependence. Both reactions proceed in low to modest efficiency (7%-44% yield) under standard reaction conditions using the E or H architectures when 10 or 20 bases separate the annealed 25 reactive groups (Figure 34, lanes 10-11 and 14-15, and Figures 35C-35D). This distance dependence limits the positions on a DNA template that can encode these or other similarly distant dependent reactions. In contrast, both 1,3-dipolar cycloaddition and reductive amination proceed efficiently (up to 97% yield) and sequence-specifically when encoded by template bases 15-25 bases away from the functionalized end of the template using the Q architecture with four 30 or five constant bases (Figure 34, lanes 12-13 and 16-17, and Figures 35C-35D). These results demonstrate that the templates 0 architecture permits distance-dependent reactions to be WO 2004/016767 PCT/US2003/025984 -121 efficiently directed by DNA bases far from the reactive end of the template. By overcoming the distance dependence of these reactions while preserving the efficiency of distant independent reactions, the. Q architecture may permit virtually any contiguous subset of bases in a single stranded 30-base template to encode any viable DNA-templated reaction. Interestingly, the Q 5 templates with only three constant bases at.their reactive ends do not consistently improve the efficiency of these reactions compared with the E-architecture (Figures 35C-35D), suggesting that four or five constant bases may be required in the £ architecture to fully realize favorable prqxiriity effects. [03681 In order to probe the structural features underlying the observed properties of the 10 Q architecture, the thermal denaturation of the 0-5 and E architectures using n =10 and n = 20 reagents were characterized. For all template-reagent combinations, only a single cooperative melting transition was observed. Compared to the E architecture reagent lacking the five-base constant region, the Q-5 reagent increased the hypochromicity upon annealing by ~-50% but did not significantly affect melting temperature in either phosphate-buffered saline (PBS) or in 50 15 mM sodium phosphate pH 7.2 with 1 M NaC1 (Figure 36). These results are consistent with a model in which template-reagent annealing in the Q architecture is dominated by coding region interactions even though the constant region forms secondary structure once the coding region is annealed. The entropic cost of partially ordering the loop between the coding and constant, regions may, therefore, be offset by the favorable interactions that arise upon annealing of the 20 constant region. 103691 DNA templates of arbitrary length are easy to synthesize and undesired cross reactivity between reactants in the same solution can be avoided using concentrations that are too low to allow non-complementary reactants to react intermolecularly. These features of DNA templated synthesis permit more than one DNA-templated reaction to take place on a single 25 template in one solution, saving the effort associated with additional DNA-templated steps and product purifications. 10370] Multiple DNA-templated reactions per step can be difficult using the E, H, or Q architectures, because the reagent oligonucleotide that remains annealed to the template following the first reaction forms a relatively rigid double helix that can prevent a second reagent 30 annealed further away along the template from encountering the reactive end of the template. To overcome this, the reactive group on the template was moved from the end of the oligonucleotide WO 2004/016767 PCT/US2003/025984 -122-, to the middle, attaching the reactive group to the'nop-Watson-Crick face of a base. This "T" architecture (see, Figure 7G) was designed to pennit two DNA-templated reactioris, one with a' reagent coupled to the 5 end of the oli gonucotide of a first transfer unit ar d one with a reagent coupled to the 3' end of the oligonucleotide of a second transfer unit, to take place sequence 5 specifically in the sane solution on a single template. [03711 TO test the 'viabilityof the T, architecture in DNA-templated reactions, the efficiency of the amine acylation, Wittig olefnation, ,3-dipolar cycloaddition, and reductive . amination reactions'using the T architecture was studied, The T architecture sequence specifically directed these four reactions with efficiencies comparable to or greater than those of 10 the E or H architectures (Figure 37, 69-100% yield when n = 1). The observed degree of distance dependence using the T architecture for each of the four reactions was consistent with the above findings (compare Figure 37 and Figure 35) Together these results dernqnstrate that the T architecture can mediate sequence-specific and efficient DNA-templated synthesis. (0372] Once the ability of the T architecture to support efficient DNA-templated, 15 synthesis was established, the ability of the T architecture to direct two DNA-templated reactions on one template in one solution was studied. Two different two-reaction schemes using the T architecture were performed. In the first scheme, depicted in Figure 38A, a benzaldehyde linked T template (1) was combined with a phospliine-linked reagent (2) and an a-iodoamide linked reagent (3) in a single solution (pH 8.5, 1 M NaC1, at 25 "C for 1 hour). The phosphine 20 linked oligonucleotide complemented ten bases of the template 5' of the aldehyde (n = -4), while the iodide-linked oligonucleotide complemented ten bases 3' of the aldehyde (n = 0). DNA templated SN 2 reaction between the phosphine and a-iodoamide generated the corresponding phosphorane, which then participated in a DNA-templated Wittig reaction to generate cinnanamide 4 in 52% overall yield after 1 hour (Figure 38B, lanes 9-10). Control reactions 25 containing sequence mismatches in either reagent generated no detectable product. The additional control reaction lacking the aldehyde group on the template generated only the SN2 reaction product (Figure 38B, lanes 3-4) while control reactions lacking either the phosphine group or the c-iodoamide group did not generate any detectable products (Figure 38B, lanes 5 8). 30 f0373] In a second two-reaction scheme mediated by the T architecture, depicted in Figure 38C, an amine-linked T template (5) was combined with a propargylglycine-linked 5' WO 2004/016767 PCT/US2003/025984 -123 reagent (6) at n -1 and a phenyl azide-linked 3' reagent (7) at n 1. The addition of 20 mM DMT-MM at pH 7.0 to induce aide forrhation followed by the addition of 500 RM copper(n) sulfate and sodium ascorbate to induce th'e recently reported Sharpless-modified Huisgen 1,3 dipolar cycloaddition provided 1,4-disubstituted triazoyl alanine adduct 8 in 32% overall yield. 5 [0374] Taken together, these observations show that the T architecture permits two sequence-specific DNA-templated reactions to take place On one template in one solution. Importantly, the T architecture templates described above were accepted as effient templatQs or.both a single cycle of primer extension as well asst indard PCR amplification using Tag DNA polymerase, consistent with the known tolerance of several DNA polymerases for 10 modifications to the non-Watson-Crick face of DNA-templates. In addition to reduding the number of separate DNA-templated steps needed to synthesize a target structure, this architecture may also permit three-component reactions commonly used to build structural complexity in synthetic libraries to be performed in a DNA-templated format. [03751 In summary, the Q and T architectures significantly expand the scope of DNA 15 templated synthesis. By enabling distance-dependent DNA-templated reactions to be encoded by bases far away from the reactive end of the template, the omega architecture expands the types of'reactions that can be encoded anywhere on a DNA template. The T architecture permits two DNA-templated reactions to take place on a single template in one step. , Materials and Methods 20 [0376] Oligonuceotide synthesis. Unless otherwise specified, DNA oligonucleotides were synthesized and functionalized as previously described using 2-(2-(4-monomethoxytrityl) aminoethoxyjethyl-(2-cyanoetbyl)-N,N-diisopropyl-phosphoramidite (Glen Research, Sterling, Virginia, USA) for 5-functionalized oligonucleotides, and using (2-dimethoxytrityloxymethyl-6 fluorenylmethoxycarbonylamino-hexane- I -succinoyl)-long chain alkylamino-CPG (Glen 25 Research, Sterling, Virginia, USA) for 3-functionalized oligonucleotides (Calderone et al. (2002) ANGEW. CHEM. INT. ED. ENGL. 41: 4104; (2002) ANGEW. CHEM. 114: 4278). In the case of templates for the T architecture, amine groups were added using 5'-dimethoxytrityl-5-JN (trifluoroacetylaminohexyl)-3-acrylimido]-2'-deoxyuridine-3'-[(2-cyanoethyl)-(N,N diisopropyl)]-phosphoramidite (Glen Research, Sterling, Virginia, USA) and then acylated as 30 reported previously (Calderone et al. (2002) supra).

WO 2004/016767 PCTf/JS2003/025984 -124 [03771 Amine Acylation. Amine-labeled and carboxylic acid-labeled DNA were combined in aqueous 1,00 mM MOPS buffet j M NaCl,TpH 7.0 (60 nM in template DNA, 120 nM in reagent DNA) in the presence of 20 mM DMT-MM> Reactions pro&beded for 12 hours at 25 0 C. 5 [03781 Wittig Olefination. Aldehyde-labe'ed and phosphorane-labeled DNA were combined iri aqueous 100,mM MOPS, 1 M NaCl, pH 7.5 (60 nM in templateIDNA, 120 nM n reagent DNA). Reactions proceeded for 2 hours at 30C. 10379] 1,3-bipolar Cycloaddition. Diabdehyde-dabeled DNA was incubated in 260 n1M N-methylhydroxylamine hydrochloride for.1 hour at room temperature (Gartner et aL (2002) J. LO AM. CHEM. Soc. 124:- 1.0304). It was subsequently combined with succinimide-labeled DNA in aqueous 50 mM MOPS, 2.8 M NaCl, pH1 7.5 final l concentrations of N-methylhydroxylamine hydrochloride 0.75 mM, 60 nM in template DNA and 90 nM in reagent DNA). Reactions proceeded for 12 hours at 37'C. [0380] Reductive Amination. Amine-labeled and aldehyde-labeled DNA were combined 15 in aqueous 100 mM MES buffer, 1 M NaCl, pH 6.0 (60 nM in template DNA, 120 nM in reagent DNA). Sodium cyanoborohydride was added as a'5 M stock in 1 M NaOH to a final concentration of 38 mM, and reactions proceeded for 2 hours at 25 *C. Reactions were quenched by ethanol precipitation in the presence of 15 rniM methylanine. [03811 TArchitecture-mediated Conversion of Compound 1 to 4. The 5'-phosphine 20 linked oligonucleotide (2) was generated by coupling N-suceinimidyliodoacetate (SIA) to the amine derived from 12-(4-monomethoxytritylamino)dodecy1-(2-yanoethy)-(NN-diisopropyl) phosphoramidite (Glen Research, Sterling, Virginia, USA) using the T (n = -4) oligonucleotide listed below, followed by treatment with 4-diphenylphosphinobenzoic acid as described previously (Garther et al. (2002) supra). The 3'-O-iodoanide-linked reagent (3) was prepared 25 by reacting the T (n = 1) oligonucleotide (see below) with SIA as described previously (Gartner et al. (2001) supra). Aldehyde-labeled template (1) was prepared by reacting the "T template" oligonucleotide (see below) withpara-fornyl benzoic acid N-hydroxysuceinimidyl ester as described previously (Gartner et al. (2002) ANGEW. CHEM. INT. ED. 41: 1796; (2002) ANGEW. CHEM. 114:1874). Template I was combined with reagents 2 and 3 in aqueous 200 mM N-(2 30 hydroxyethyl)piperazine-N'-(2-ethanesulfonic acid) (HEPES) buffer at pH 8.5 with 1 M NaCl, (63 nM template and 125 nM of each reagent). Reactions proceeded for up to 1 hour at 25 *C.

WO 2004/016767 PCT/US2003/025984 -125 [0382) The results of denaturing polyacrylajnide gel electrophoresig analysis of these reactions is shown in Figure 38B. The 30-base T architecture template (1) containingan aldehyde group was present in lanes 1 -2 and lanes 5-10. A template lackiri he aldehyde group but otherwise identical to (1) was present in lanes 3 and 4. DNA-linked phosphine reagent (2) 5 was present in lanes 3-6 and lanes 9-10. -DNA-linked a-iodoamide reagent (3) was present in lanes 3-4 and lahes 7-10. Lanes 1, 3, 5, 7, and 9 shoW. reactions after 30 minutes. Lanes 2, , 6, 8, and 10 show reactions after 1 hour. [0383] T A rchitecture-mediated Con version of Compound 5 to 8. The 5' propargylglycine linked oligonucleotide (6) was generated by combining the corresponding T (n 10 -1) 5'-amine-linked reagent oligonucleotide (see belowy) with 2 mg/mL bis(sulfosuccinimidyl)suberate in 9:1 200 mM sodium phosphate pH 7.2:DMF for 10 minutes at 25 'C, followed by treatment with 0.3 vol of 300:mM racemic propargyiglycine in 3Q0 mM NaOH for 2 hours at 25 *C. The Y-azido linked oligonucleotide (7) was generated by combining the T (n = 1) amine-linked reagent oligonucleofide (see below) with 2 mg/mL (N 15 hydroxysuccinimidyl)-4-azidobenzoate in 9:1 200 mM sodium phosphate pH 7.2:DMF for 2 hours at 25 'C. Reagents 6 and 7 were purified by gel filtration and reverse-phase HPLC. Template 5 and reagents 6 and 7 were combined in aqueous 100 mM MOPS pH 7.0 in the . presence of 1 M NaCl and 20 mM DMT-MM for 12 hours (60 nM template, 120 nM reagents) at 25 'C. Copper (II) sulfate pentahydrate and sodium ascorbate were then added to 500 RM each. 20 After 1 hour at 25 *C, reactions were quenched by ethanol precipitation. [0384] DNA Oligonucleotide Sequences Used. E or Q template: 5'-H 2 N-GGT ACG AAT TCG ACT CGG GAA TAC CAC CTT [SEQ ID NO: 58]. H template: 5'-H 2 N-CGC GAG CGT ACG CTC GCG GGT ACG AAT TCG ACT CGG GAA TAC CAC CTT [SEQ ID NO: 59]. T template: 5'- GGT ACG AAT TCG AC(dT-NH 2 ) CGG GAA TAC CAC CTT 25 [SEQ ID NO: 60]. E or H reagent (n = 1): 5'-AAT TCG TAC C-NH 2 [SEQ ID NO: 61]. E or H reagent (n = 10): 5'-TCC CGA GTC G-NH 2 [SEQ ID NO: 62]. E or H reagent (n 20): 5' AAG GTG GTA T-NH 2 [SEQ ID NO: 63]. Mismatched E or H reagent: 5'-TCC CTG ATC G

NH

2 [SEQ ID NO: 64}. Q-3 reagent (n = 10): 5'-TCC CGA GTC GAC C-NH 2 [SEQ ID NO: 65]. Q-4 reagent (n = 10): 5'-TCC CGA GTC GTA CC-NH 2 [SEQ ID NO: 66]. n-5 reagent (n 30 = 10): 5'-TCC CGA GTC GGT ACC-NH 2 [SEQ ID NO: 67]. K-3 reagent (n =20): 5'-AAG GTG GTA TAC C-NH 2 [SEQ TD NO: 68]. 0-4 reagent (n = 20): 5'-AAG GTG GTA TTA CC- WO 2004/016767 PCTiUS2003/025984 -126 NH2 [SEQ ID NO: 69]. 0-5 reagent (n 20): 5'-AAG GTG.GTA TGT ACC- NH2'[SEQ ID NO: 70]. Mismatched 0-3 reagent: 5'-TCC CTG ATC GAC C-NH 2 [SEQ ID NO: 71]. Mismatched 0-4 reagent: 5'-TCC CTG ATC GTA CC-NH 2 [SEQ ID NO: 721. Mismatched i 5 reagent: 5'-TCC CTG ATC GGT ACC-NH2 [SEQ ID NO: 73]. T reagent (n = 1): 5'-GGT 5 ATT CCC G-NH 2 [SEQ ID NO: 74], T reagent (n = 2): 5'-TGG TAT TCC C-NHz [SEQ ID NO: 75]. T reagent (n = 3): 5'-GTG GTA TTC C-NH-I 2 [SEQ ID NO: 76] T reagent, (n = 4): 5'-GGT GGT ATT C-NH 2 [SEQ ID NO: 77]. T reagent (n = 5): 5'-AGG TGG TAT T-NH2 SEQ ID NO: 78]. T reagent (n = -1): 5'-NH 2 -GTC GAA TTC G [SEQ ID NO: 79]. T reagent (n -4) for 2: 5'-[Cu-amine linker]-AAT TCG TAC C [SEQ ID NO: 80]. 10 [0385] Reaction yields were quantitated by denaturing polyacrylamide gel electrophoresis followed by ethidium bromide staining, UV visualization, and CCD-based densitometry of product and template starting material bands. Yield calculations assumed that templates and products were denatured and, therefore, stained with comparable intensity per base; for those cases in which products are partially double-stranded during quantitation, changes 15 in staining intensity may result in higher apparent yields. Representative reaction products were characterized by MALDI mass spectrometry in addition to denaturing polyacrylamide gel electrophoresis. 10386] Melting curves were obtained on a Hewlett-Packard 8453 UV-visible spectrophotometer using a Hewlett-Packard 89090A Peltier thermocontroller. Absorbances of 20 template-reagent pairs (1.5 pM each) at 260 nm were measured every 1 "C from 20 'C to 80 *C holding for 1 minute at each temperature in either phosphate-buffered saline ("PBS," 137 mM NaCl, 2.7 mM potassium chloride, 1.4 mM potassium phosphate, 10 mM sodium phosphate, pH 7.4) or in high salt phosphate buffer ("HSB," 50 mM sodium phosphate pH 7.2, 1 M NaCl). Example 6: Stereoselectivity in Nucleic Acid-Templated Synthesis 25 [03871 This Example demonstrates that it is possible to perform stereoselective nucleic acid-templated syntheses. The chiral nature of DNA raises the possibility that DNA-templated synthesis can proceed stereoselectively without the assistance of chiral groups beyond those present in DNA, thereby transferring not only sequence but also stereochemical information from the template to the product.

WO 2004/016767 PCT/US2003/025984 -127 103881 Steteoselectivity was examined in tle context of DNA-templated nucleoplhilic substitution reactions. ;Hairpin architecture templates cojugated at their 5' amino'terrnifi directly to (S)- or (R52-bromopropionamide :were combined with 3' thiol-lihked reagent oligonucleotides at 25, C (Figure 39A) (Gartnr et -al. (2001) supra; Gartner et al. (2003) 5 ANGEW. ChEM. INT. ED. 42: 1370). The exact structure of the hairpin template and its complimentary reagent (Figure 39A) Were asTollows: Template: 5'-BrCH(CH 3 )CONIC-TQG CGA GCG TAC GCT CCC GAG OTA. CGA ATT C-3' [SEQ ID NO: 81]1 Reagent: 5 t -GAA TTC GTA CC-(CEI 2

)

3 SH-31 [SEQ ID NO: 82] 10 [03891 The stability of the bromides under the reaction conditions was confirmed by several independent methods. Initial rates of thioether product formation were determined by denaturing gel electrophoresis and the product's Were additionally characterized by MALDI-TOF mass spectrometry. Apparent rates of product.formation were 4.0±0.2-fold higher for (S) bromide-linked templates than for (R)-bromide-linked templates. Because template-reagent 15 annealing could be partially rate-determining, this value is a lower limit of the actual ratio of ks/kR, assuming annealing rates are unaffected by bromide stereochemistry. [0390] Surprisingly, similar preferences favoring the (S)-bromide were also observed using end-of-helix template architectures (Figure 39B), even when 12 nucleotides separated the thiol and bromide in the template-reagent complexes. The exact structure of the end-of-helix 20 template and its complimentary reagent (Figure 39B) were as follows: Template: 5'- BrCH(CH 3 )CONH-TAC GCT CGC GAT GGT ACG AAT TC-3' [SEQ ID NO: 83] Reagent: 5'-GAA TTC GTA CC-(CH 2

)

3 SH9-3' [03911 Stereoselectivity appeared to be independent of whether the bromide or the thiol 25 was conjugated to the template (Figures 39B and 39C). The exact structure of the end-of-helix template conjugated to the thiol and its complimentary reagent (Figure 39C) were as follows: Template: 5'-GAA TTC GTA CAT AGC GCT CGC AT-(CH 2

)

3 SH-3' [SEQ ID NO: 84] Reagent: 5'- BrCH(CH3)CONH-TGT ACG AAT TC-3' [SEQ ID NO: 85] WO 2004/016767 PCT/JS2003/025984 - 128 [0392] Similar selectivities emerged from pseudo-kinetic resolutions containing both bromide stereoisomers in which thioether'products arising froin (S)- and (R)-bromides were distinguished using templates of two distinct lengths (ks/k'= 4.2±0.4 to 4.9±0.3). Taken' together, these findings indicate that the chirality of a DNA template can be. transferred to 5 products of DNA-templated syntliesis that do not resemble the DNA backbone. 103931 In order to probe the origins of the observed stereoselectivity, a series of template and reagent analogs were synthesized in which nucleotides near the thiol or bromide were i-cplaecd with flexible achiral linkers. Replacing the12 -template nucleotides separating the bromide and thiol in either of the end-of-helix reactions with an achiral polyethylene glycol 10 linker of similar length (72 bonds) resulted in the loss of stereoselectivity. Stereosefectivity was also-abolished when flexible achiral linkers consisting of three or five consecutive methylene or ether oxygens were inserted between the 5'end of the template oligonucleotide and the thiol or bromide groups, or between the 3' end of the reagent bligonucleotide and the thiol or bromide. Chiral linkers between reactants, therefore, are required for stereo selectivity in this DNA 15 templated reaction. These results also suggest that both the thiol and the bromide participate in the rate-determining step of the reaction, consistent with an SN 2 mechanism. [0394] The known sensitivity of single- and double-stranded DNA conformations on distal base stacking or base pairing interactions suggests that groups distal from the bromide or thiol could play important roles in inducing stereoselectivity. To test these possibilities, 11 of 20 the 12 template nucleotides closest to the 5' bromide were replaced in the end-of-helix reaction with chiral abasic phosphoribose linkers in which the aromatic base was replaced with a proton (Figure 40A). The exact structure of the end-of-helix template was the same as in Figure 39, except that bases 2-12 were replaced with abasic phosphoribose units (prepared from the corresponding phosphoramidite from Glen Research, Sterling, Virginia, USA). Even though the 25 5' thymidine nucleotide closest to the bromide was unchanged, the resulting reactions were not stereoselective, indicating that the nucleotide closest to the bromide was not sufficient to induce the observed stereoselectivity. 103951 Each of the 11 missing aromatic bases from the 5' end were then restored (Figure 40B) and measured rates of (S)-bromide and (R)-bromide reaction for each resulting template. 30 Surprisingly, no stereoselectivity was observed when up to five bases were restored. Stereoselectivity increased steadily up to ks/kR= 4.3 when 6 through 11 bases were restored WO 20041016767 PCT/US2003/025984 -129 (Figure 40C). Restoration of the missing aromatic bases from the 3' end of the abasic region instead of from the 5' end also induced stbreoselectivity only after several bases were restored (five t6 11 bases in this case) (Figure 401j). Collectively, these findings suggest that stereo selectivity arises from the conformation of nucleotides adjacent to either reactant, and that 5 the conformation(s) leading to stereoselectivity require at least 5-6 consecutive aromatic bases. t0396] This model of stereoselectivity predicts that global conformational changes in the template-reagent complex may alter stereoselectivity even if the covalent structure and absolute stereochemistry of all reactants were preserved. Double-stranded DNA sequences rich in (5-Me C)G repeats can adopt a left-handed helix (Z-form) rather than the usual right-handed helix (B 10 form) at high salt concentrations (Rich et at. (1984) J. ANNU. REv. BIOCHEM. 53: 791-846; Behe etal. (1981) PROC. NATL. ACAD. SC. USA 78: 1619-1623; Mao et al. (1999) NATURE 397: 144 146). Bromide-linked (5-Me-C)G-rich hairpin templates and complementary thiol-linked reagents protected as unreactive disulfides were prepared. When combined in equimolar ratios, the circular dichroism (CD) spectra of the resulting template-reagent complexes in low salt (100 15 mM NaCl) were characteristic of B-form DNA (see, for example, Figure 42D). In the presence of high salt concentrations (5 M NaCl or 2.5 M Na 2 S04), the same template-reagent complexes exhibited CD spectra representative of Z-form DNA. In contrast, the CD spectra of template reagent complexes of normal sequence were representative of B-form DNA under both low salt and high salt conditions (see, for example, Figure 42C). 20 [0397] The stereoselectivity of DNA-templated reactions between bromide-linked templates and thiol-linked reagents using either the mixed or (5-Me-C)G-rich sequences was examined in the presence of low or high salt concentrations. The mixed sequence templates and reagents (B-form DNA) in the presence of low or high salt concentrations favored the (S) bromide by 4.3- or 3.2-fold, respectively (Figure 41A). The (5-Me-C)G-rich template and 25 reagent in low salt concentrations (B-form DNA) exhibited a 4.4-fold preference for reaction of the (S)-bromide (Figure 41 A). Remarkably, repeating this reaction in the presence of high salt concentrations that induce Z-form DNA resulted in a 14-fold change in stereoselectivity now favoring the (R)-bromide by 32-fold (ks/ka= 0.31) (Figure 41B). This inversion of stereoselectivity as a result of changing the handedness of the DNA double helix is consistent 30 with the theory implicating the conformation of the template and reagent in determining the stereo selectivity of this DNA-templated reaction.

WO 2004/016767 PCTUS2003/025984 -' - 130' 103981 These experiments demonstrate tha stereos-lectivity can be imparted during nucleic adid -templted organic synthesis. Cohformatiohs of DNA dependent on base stacking together with a patiIally constrained resenfation of reagtants appear to be responsible for the observed stereoselectivity. These experiments further demonstratee that'a single structure with 5 one absolute stereocheinistry can induce opposite stereoselectivities when its macromolecular conformation ig altered. . Oligon cleotides . {03991 Th6 exact structures of the ternplates contajiing mixed and (5-Me-C)G-rich sequence, and their corresponding reagents used; are as follows: 10 Mixed sequence: Template: 5'-GAA TTC TGG ACA CTT. AGC TAT TCA TCG AGC GTA CGC TCG ATG AAT AGCZ(CH 2

)

3 SH-3' [SEQ ID NO: 861 Reagent: 5'-BrCH(CH-3)CONH-TAA GTG TCC AGA ATT C-3'[SEQ ID NO: 87] (5-Me-C)G-rich sequence: 15 Template:' 5'-GAA TTC C*GC* GC*G C*GC* AC*G C*GC* GC*G C*GG AGC GTA CGC TCC* GC*G C*GC* GC*G-(CH2)3SH-3' [SEQ ID NO: 88] Reagent: 5'- BrCH(CH 3 )CONH-TGC* GC*G C*GC* GGA ATT-3' [SEQ ID NO: 89] C*= 5-methyl cytosine. The thiols in both the mixed and (5-Me-C)G-rich sequences 20 were protected as disulfides (-(CH 2

)

3

S-S(CH

2

)

3 0H) fdr circular dichroism measurements. DNA Synthesis and Analysis [0400j DNA oligonucleotides were synthesized on a PerSeptive Biosystems Expedite 8090 DNA synthesizer using standard phosphoramidite protocols and were purified by reverse phase HPLC with a triethylammonium acetate (TEAA)/CH 3 CN gradient. Oligonucleotides were 25 quantitated by UV and by denaturing PAGE after staining with ethidium bromide. Quantitation of DNA by denaturing PAGE was performed with a Stratagene Eagle Eye II densitometer. Synthetically modified oligonucleotide analogs were incorporated using the corresponding phosphoramidites or controlled pore glass (CPG) beads purchased from Glen Research, Sterling, Virginia, USA.

WO 2004/016767 PCT/US2003/025984 - 131 DNA Functionalization [0401] 2-bromopropionamide-NlIS esters. 200 mg N-hydroxysuccinimide (Pierce Rockford, IL, USA) was dissolved in mnhydrous CH 2 Cl 2 together with 1.1 equivalehts of'a 2 bromopropionic acid (either racemic, (R)-, or (S)-) and 2 equivalents of 1-(3 5 dimethylaminopropyl)-3-ethylcarbddiimide (EDC) (Aldrich). The 2-bromopropionic acid enantiomers were >95% enantiopure as judged by chiral HPLC (5% isopropanol in hexanes, (R,R) WHELK 01 chiral phase, detection at 220 nm)I The reaction was maintained at room temperature and complete after 1.5 hours as judged by TLC (EtOAc). The crude reaction mixture was extracted with 2.5% sodium hydrogen sulfate (NaHSO 4 ) to remove the excess EDC. 10 The organic phase was washed with brine, dried over magnesium sulfate (MgSO 4 ), and concentrated in vacuo. The residue was dried and used directly for DNA functionalization. 104021 5'-functionalization of oligonucleotides. An NHS ester prepared as described above was dissolved in DMSO. Up to Io tg of a 5"amino DNA oligonucleotide was combined with 3 mg/mL NHS ester (final reaction = 10% DMSO) in 200 mM sodium -phosphate 15 (pH = 7.2) at room temperature for 2 hours. The functionalized oligonucleotides were purified by gel filtration and reverse-phase HPLC, and were characterized by denaturing PAGE and MALDI-TOF mass spectrometry. (0403] 3'-thiol modified oligonucleotides. The Y thiol group was incorporated by standard automated DNA synthesis using 3'-disulfide-linked CPG (Glen Research, Sterling, 20 Virginia, USA). Following oligonucleotide synthesis, the disulfide was cleaved with 50 mM DTT, IM TAPS (pH -8.0) at room temperature for 1 hour and purified by gel filtration before being used in DNA-templated reactions. DNA-templated Reactions {04041 Reactions were performed with 60 nM template and 60 nM reagent in 50 mM 25 MOPS (pH = 7.5) and 250 mM NaCI at 25 'C unless otherwise specified. Reaction aliquots were removed at time points from 2 minutes to 120 minutes and quenched with excess p.

mercaptoethanol. Starting materials and products were ethanol-precipitated from the quenched reaction mixtures, analyzed by denaturing PAGE, quantified as described above. Relative initial rates of product formation were determined from the fitting the raw yield vs. time data and were 30 used to calculate ks/kR. Representative data are shown in Figure 42.

WO 2004/016767 PCT/US2003/025984 - 132 [04051 For the representative data sets shoWn in.Figure 42, the apparent second order rate constants derived from the initial rates are psifollows:. [04061 Figures 39A and 42A: kR,ap .1.94 x 1i0 M's'; ks,app 7.07 x 103 M s; k,pp = 4.58 x 10 MWsW 5 104071 Figures 39B and 42B: kR,app, 5.83 x 103 M's-; ks,app= 21.9k0 3 Ms kre,app= 13.6 x 10' M- s [0408] Figures 42C and 44A, low salt: ka,app -.4.00-x 10 M ks,app = 17.6 x 10 M- I;kracapp = 9.88 x 103 M- S [0409] Figures 42C and 44A, high salt; 10 kapp- 5.95 x 10' M's-; ks,app =18 M's'; kracapp= 10.8 x i03 Ms' [04101 Figures 42D and 44B, low salt: kR,pp =6.11 x 10' Ms; ks,app = 25.4 x 103 M-1s-; kac,app 12.1 x 10 M's' {0411] Figures 42D and 44B, high salt: kR,app = 24.6 x 10 Ms"; ksapp = 7'66 x 10' M-Ys; kracapp 13.6 x 103 M-S 15 Evaluating Bromide Stability [0412] The structural and configurational stability of the bromides under the reaction conditions was confirmed by several independent methods. Each bromide-linked template or reagent oligonucleotide was pre-incubated for up to 72 hours at 25 0 C, and up to 48 hours at 37*C under the reaction conditions in the absence of thiol. Following the pre-incubation, 20 stereoselectivity was measured as described above and always found to be unchanged as a result of the pre-incubation. In addition, large-scale (250 pmol) quantities of bromide-linked templates ((R), (S), and pseudo-racemic) were each incubated under the reaction conditions for 16 hours and analyzed by MALDI-TOF mass spectrometry. No evidence of bromide displacement (by water or by chloride) was observed as shown in Tables 11 and 12. 25 TABLE 11: End-of-helix template (expected mass 7202. 1) Isomier ,Observed Mass (R) bromide: before incubation = 7203.3+7 WO 2004/016767 PCT/IUS2003/025984 -133 after incubation =7206.4±7 (S) bromide: before incubation =7206.0&7 after iricub ation = 7201 9t7 (+) bromide: mass before incubation -7201.7±7 rhass after incubation =7204-717 TABLE 12: H airpin template (expected mass 9682.4) , ~~~Isomnr.'be:e Ms (R) bromide: mass before incubation = 9686 6 10 mass after incubation = 9685.7±10 (S) bromide: mass before incubation = 9683.8±10 mass after incubation = 9680.6+10 (I) bromide: mass before incubation = 9680.6+10 mass after incubation = 9684.710 10413] Finally, small molecule analogs of the above bromide-linked DNAs (both 5 enantiomers of N-methyl 2-bromopropionamide) were incubated for 16 hours under the reaction conditions and analyzed by chiral HPLC under conditions that resolve the (S)- and (R) enantiomers. No change in retention time was observed. Stereoselectivities Using Achiral Flexible Linkers [04141 Figure 43 shows modified template or reagent structures that result in loss of 0 stereoselectivity during DNA-templated SN 2 reactions. In all cases, kSapp R,app values fell within the range of 0.95 to 1.09 (+0.09), which reflects the mean and standard deviation of at least three independent experiments. The exact structures of the templates containing achiral linkers and their corresponding reagents were as follows: [0415] Figure 43A: 5 Template 5'-BrCH(CH3)CONH-[(CH 2

)

2 0 2 OP0 3 - ([(CH 2

)

2 0]OOPO} 3 -GGT ACG AAT TC-3' [SEQ ID NO: 90] WO 2004/016767 PCT/US20031025984 -134 Reagent;-- 5'-GAA TTC GTA CC-(CH) 3 SH-3 [SEQ ID NO: '911 10416] Figure 43B: Template: 5'-GAA IFTC GTA CA-(CH2) 3 'OPO3-{[(CH2)201OPO3 }3(CH2)3SH-3 [SEQ ID NO: 92] 5 Reagent 5'. BrCH(CI3)CONH-TGT ACG AAT TC-3' [SEQ ID NO:.93] [04171 Figure.43C: Template: 5'- BrCH(CH3)CONH-(CI-)2q]2OPO3--AC GCT CGC GAT GGT ACO AAT TC-3' [SEQ ID NO:'941 Reagent:. 55-GAA TTC GTA CC-(CH2) 3 SH-3' [SEQ ID NO: 95] 10 [0418] Figure 43D: Template: 5'-GAA TTC GTA CAT AGC GCT CGC A-(CH) 3 OPO3 -(CH 2 )3SH-3 [SPQ ID NO: 96] Reagent: 5'- BrCH(CH)CONH-TGT ACG AAT TC-3' [SEQ ID NO: 97] 10419] Figure 43E: 15 Template: 5'- BrCH(CH3)CONH-TAC GCT CGC OAT GGT ACG AAT TC-3' [SEQ ID NO: 98] Reagent; 5-GAA TTC GTA CC-(CH2) 3 0PO3-(CH2)3SH-3' [SEQ ID NO: 99] [0420] Figure 43F: Template: 5'-GAA TTC GTA CAT AGC GCT CGC AT-(CH2) 3 SH-3' [SEQ ID NO: 20 100] Reagent: 5'-BrCH(CH 3 )CONIH-[(CH2)20]2OPO3~-TGT ACG AAT TC-' [SEQ ID NO: 101] Circular Dichroism (CD) of B-DNA and Z-DNA [0421] The DNA templates and reagents were prepared as described above. Thiol-linked 25 reagents were not deprotected and remained in their disulfide forms during CD analysis. CD samples contained 215 nM template and 215 nM protected reagent in 50 mM phosphate buffer (pH = 7.5) with either 100 mM or 5 M NaCl. A background sample lacking DNA was also WO 2004/016767 PCT/US2003/025984 -135 prepared for each sample. The CD-measurements vere performed in a I mm path euvette at 25 0 C scanning from 3,60 nm to 200 nmat 2unr/see on a JASCO polarized spectrometer with a 2.0 nm resolution. The resulting CD spectra'of B-form and Z-form tempiate-reagent complexes are shown in Figure 44: Figure 44A shows circular dicliroism (CD) spectra of template-reagent 5 complexes'containing normal (mixed'composition) sequences which are characteristic. qf B DNA. Figure 44B shows CD spectra of (5-Me-C)G-rich complexes having a B-DNA conformation at low salt concentrations, and haing a'Z-DNA conformation at high salt concentrations. The exact-structures of the templates containing mixed and (5-Me-C)G-rich sequence, and tieir corresponding reagents used, are as follows: 10 [04221 Mixed sequence: Template: 5'-GAA TTC TGG ACA CTT AGC TAT TCA TCG AGC GTA CGC TCG ATG AAT AGC-(CH2)3SI-3' (SEQ ID NO: 102] 1 (The thiol was protected as a disulfide [(C1 2

)

3

S-S(CH

2

)

3 0H]'for circular dichroism measurements), 15 Reagent: 5'-BrCH(CH 3 )CONH-TAA GTG TCC AGA ATT C-3' [SEQ ID NO: 103 104231 (5-Me-)G-rich sequence: Template: 5'-GAA TTC C*GC* GC*G C*GC* AC*G C*GC* GC*G C*GG AGC GTA CGC TCC* GC*G C*GC* GC*G-(CH 2 )3SH-3' [SEQ ID NO: 104] (The thiol was protected as a disulfide [(CH 2

)

3

S-S(CH

2

)

3 OHlfor circular 20 dichroism measurements) Reagent: 5'- BrCH(CH3)CONH-TGC* GC*G C*GC* GGA ATT-3' [SEQ ID NO: 105] C* 5-methyl cytosine Stereoselectivity Induced by 1-form and Z-form DNA 25 [04241 Figure 45 shows a representative denaturing gel electrophoresis analysis of reactions using the CG-rich sequences at 100 mM NaCl (lanes 1-3) or at 5 M NaCl (lanes 4-6) (6 hour time point). Lanes I and 4: racernic bromide; lanes 2 and 5: (R)-bromide; lanes 3 and 6: (S)-bromide. The bromide-linked reagent is not visible. Similar results were observed using Na 2

SO

4 ind of NaCi.

WO 2004/0.16767 PCT/US2003/025984 -136 DNA-templated Reactions in the Presence of NaIS0 4 instead of NaCl 10425] In order to ascertain that the observed stercoselectivities were not affected by the presence of chloride, the experiments shown in Figures 39 and 44 were repeated in the presence of Na 2

SO

4 instead of NaCl (keeping the concentration of sodium constant). The results of three. 5 independent trials were very similar to those reported in the presence of NaCl, and are as follows: {0426] Figure 39A with Na 2

SO

4 instead of NaCI: ks/kR 5.4 0.5 104271 Figure 39B witi Na 2

SO

4 instead of NaCl: ks/kR =,3.9 + 0.3 [0428] Figure 39C with Na 2

SO

4 instead of NaCI: ks/kR 4.7 ± 0.7 10 {0429] Figure 44A, low salt with Na 2

SO

4 instead of NaCl: ks/kR= 3.7 i 0.7 10430] Figure 44A, high salt with Na 2

SO

4 instead of NaCL: ks/kR = 3.1 0.6 10431] Figure 44B, low salt with Na 2

SO

4 instead of NaCl: ks/kR = 3.6 ± 0.5 (0432] Figure 44B, high salt with Na 2

SO

4 instead of NaCl: ks/kR =0.25 ± 0.03 MALDI-T OF Mass Spectrometry of Representative Products 15 104331 - The products from the representative DNA-templated reactions (240 pmol scale) in Figure 39 were purified by preparative denaturing polyacrylamide gel electrophoresis followed by extraction with 0.1 M triethylammonium acetate at 37 "C overnight. The lyophilized products were subjected to MALDI-TOF mass spectrometry, the results of which are summarized in Table 13. In all cases the observed mass is consistent with the expected mass. 20 TABLE 13 Figure Expected Mass Observed Mass 39A 13067.5 13015.6+65 39B 10562.0 10587.2±53 39C 10558.1 10600.1±53 WO 2004/016767 PCT/US2003/025984 -137 Example 7: Directing Otherwise Incompatible Reactions in a Single Solution [0434] This Example demonstrates iat oligonucleotides can simultaneously direct several different synthetic reaction types Within the same solution, even though the reactants involved would be cross-reactive and, therefore, incompatible under traditional synthesis 5 conditions: These findings also demonstrate that it is possible to perform a one-pot diversification "of synthetic library precursors into products using multiple, simultaneous and not necessarily compatible reaction types. 104351 The ability of DNA template tomediat-diversification using different reaction types without spatial separation was -initially tested by preparing three oligonucleotide templates 10 of different DNA sequences (1a-3a) functionalized at their 5' ends with maleimide groups and three oligonucleotide reagents (4a-6a) functionalized at their 3' ends with an amine, thiol, or nitroalkane group, respectively (Figure 46). The DNA sequences of the three reagents each contained a different 10-base annealing region that was complementary to ten base near the 5' end of eabh of the templates. Combining la with 4a, 2a With 5a, or 3a with 6a in threeseparate 15 vessels at pH 8..0 resulted in the expected DNA-templated amine conjugate addition, thiol, conjugate addition, or nitro-Michael addition products 7-9 (Figure 46, lanes 1-3). [04361 To distinguish the nine possible reaction products that could be generated upon combining l a-6a, the lengths of template oligonucleotides were varied to include 11, 17, or 23 bases and the lengths of reagent oligonucleotides were varied to include 14, 16, or 18 bases. 20 Differences in oligonucleotide length were achieved using extensions distal from the reactive groups that did not significantly affect the efficiency of DNA-ternplated reactions. This design permitted all nine possible reaction products (linked to 25, 27, 29, 31, 33, 35, 37, 39, or 41 bases of DNA) to be distinguished by denaturing polyacrylamide gel electrophoresis. {0437] A solution containing all three templates (la-3a) was combined with a solution 25 containing all three reagents (4a-6a) at pH 8.0. The resulting reaction exclusively generated the three desired products 7, 8, and 9 of lengths 25, 33, and 41 bases indicating that only the three reactions corresponding to the complementary template-reagent pairs took place (Figure 46, lane 4). Formation of the other six possible reaction products was not detected by densitometry (<5% reaction). In contrast, individually reacting templates and reagents containing the same, rather 30 than different, 10-base annealing regions permitted the formation of all possible products (Figure 46, lane 5). This result demonstrates the ability of DNA-templated synthesis to direct WO 2004/016767 PCT/US2003/025984 -138 the selective one-pot transformation of a single functional group into three distinct types of products (in this Example, maleimide intd secondary amine, thioether, or a-branched nitro alkane). {04381 To test the ability of this diversification mode to support one-pot reactions 5 requiring n6n-DNA-linked accessory reagents, an analogous experiment was conducted with two aldehyde-linked reagents either 14 or 16 bases in length (4b or Sb, respectively) and a complementary 11-base amine-linked template (1b) or a 17-base phosphorane-liiked template (2b): Combining lb and 4b at pH 8.0 inthe presence of 3 mM NaBH 3 CN resulted in the DNA templated reductive amination product 10, while 2b and 5b under the same conditions generated 10 Wittig olefination product 11 (Figure 46). Mixing all four reactants together in one'pot resulted in an identical product distribution as the combined individual Wittig olefination or reductive amination reactions (Figure 46): No reaction between amine lb and aldehyde 5b or between phosphorane 2b and aldehyde 4b was detected (Figure 46, lane 8 versus lane 9). [04391 The generality of this approach was explored by including multiple reaction types 15 that required different accessory reagents. Three amine-linked templates (1c-3c) of length 11, 17, or 23 bases were combined with an aldehyde-, carboxylic acid-, or maleimide-linked reagent (4c-6c)'14, 16, or 18 bases in length, respectively, at pH 8.0 in the presence of 3 mM NaBH 3 CN, 10 mM 1-(3-dimethyl-aminopropyI)-3-ethylcarbodiimide (EDC), and 7.5 mM N- , hydroxylsulfosuccinimide (sulfo-NHS). The reactions containing all six reactants afforded the 20 same three reductive amination, amine acylation, or conjugate addition products (12-14) that were generated from the individual reactions containing one template and one reagent and did not produce detectable quantities of the six possible undesired products arising from non-DNA templated reactions (Figure 46, lanes 10-14). Collectively, these results indicate that DNA templated synthesis can direct simultaneous reactions between several mutually cross-reactive 25 groups in a single pot to yield only the sequence-programmed subset of many possible products. [0440) The above three examples each diversified a single functional group (maleimide, aldehyde, or amine) into products of different reaction types. A more general format for the one pot diversification of a DNA-templated synthetic library into products of multiple reaction types would involve the simultaneous reaction of different functional groups linked to both reagents 30 and templates. To examine this possibility, six DNA-linked nucleophile templates (15-20) and six DNA-linked electrophile reagents (21-25) collectively encompassing all of the functional WO 2004/016767 PCT/US2003/025984 ' -139 groups used in the above three examples (amnine,laldehyde; maleimide, carboxylic acid, nitroalkane, yhosphorane and thiol)"were preparea.(Figtire 47). These twelve DNA-linkad reactants could, in theory, undergo simniltanepus arnine.cpnjugate additionjthiol conjugate addition, nitro-Michael addition, reductive amination, amine acylation, and Wittig olefination in 5 the same pdt, although the apparent second order rate constants of these six reactionss vaxy by more than .10-fold. 10441] Detennining the outcome of combining all twelve reagents and templates in a single pot by using oligonucleotides of varying lengths-ji difficult due the large number (at least 28) of possible products that could be generated. Accordingly, the length of the reagents 'as 15, 10 20, 25, 30, 35, or 40 bases were varied but the length of the templates was fixed at 11 bases (Figure 47). Each of the six complementary ierplate-reagent pairs when reacted separately at pH 8.0 in the presence of 3 mM NaBH 3 CN; 10 md EC, and 7.5 mM sulfo-NHS generated the expected amine'conjugate addition, thiol conjugate addition, nitro-Michael addition, reductive amination, amine acylation, or Wittig olefinafion products (Figure 47). Reaction efficiencies 15 were greater than 50% relative to the corresponding individual reactions despite having to compromise between differing optimal reaction conditions. Templates 15-20 were also prepared in a 3'-biotinylated form. The biotinylated templates demonstrated reactivities indistinguishable from those of their non-biotinylated counterparts (Figure 47). [0442] Six separate reactions each containing twelve reactants then were performed at 20 pH 8.0 in the presence of 3 mM NaBH 3 CN, 10 mM EDC, and 7.5 mM sulfo-NHS (Figure 48). Each reaction contained a different biotinylated template (15, 16, 17, 18, 19, or 20) together with five non-biotinylated templates (from 15-20) and six reagents (21-25). These reactions were initiated by combining a solution containing 15-20 with a solution containing 21-25. The products that arose from each biotinylated template were captured with streptavidin-coated 25 magnetic beads and identified by denaturing gel electrophoresis. Because the six reagents in each reaction contained oligonucleotides of unique lengths, the formation of any reaction products involving the biotinylated templates and any of the reagents could be detected. In all six cases, the biotinylated template formed only the single product programmed by its DNA sequence (Figure 48) despite the possibility of forming up to five other products in each 30 reaction. Taken together, these findings indicate that reactions of significantly different rates requiring a variety of non-DNA-linked accessory reagents can be directed by DNA-templated WO 2004/016767 PCT/US2003/025984 -140 synth6sis in the same solution, even when both templates and reagents contain several different crossreactive functional groups. The ability of DNA templates to direct multiple reactions at concentrations that exclude non-templated reactions from proceeding at appreciable rates' mimics, in a single solution, a spatially separated set of reactions. 5 [04431 Compared to the use of traditional synthetic methods, generating libraries of small molecules by DNA-templated synthesis is limited by several factors including the need to prepare DNA-linked reagents, the restriction of aqueous, DNA-compatible chemistries; and the reliance on characterization methods such as mass spectrometry and electrophoresis that are appropriate for molecular biology-scale (pg to pg) reactions. On the other hand, DNA-templated 10 synthesis (i) allows the direct in vitro selection (as opposed to screening) and amplification of synthetic molecules with desired properties, (ii) permits the preparation of synthetic libraries of unprecedented diversity, and (iii) requires only minute quantities of material for selection and identification of active library members. In addition, this Example demonstrates that potentially useful modes of reactivity not possible using current synthetic methods can be achieved in a 15 DNA-templated format. For example, six different types of reactions can be performed simultaneously in one solution, provided that required non-DNA-linked accessory reagents are compatible. This reaction mode permits the diversification of synthetic small molecule libraries using different reaction types in a single solution. Materials and Methods 20 Synthesis of Templates and Reagents [0444] Oligonucleotides were synthesized using standard automated solid-phase techniques. Modified phosphoranidites and controlled-pore glass supports were obtained from Glen Research, Sterling, Virginia, USA. Unless otherwise noted, functionalized templates and reagents were synthesized by reacting 5'-H 2

N(CH

2 0) 2 terminated oligonucleotides (for 25 templates) or 3'-OP0 3

-CH

2

CH(CH

2 OH)(CH2)4TH2 terminated oligonucleotides (for reagents) in a 9:1 mixture of aqueous 200 mM pH 7.2 sodium phosphate buffer:DMF containing 2 mg/mL of the appropriate N-hydroxysuccinimide ester (Pierce, Rockford, IL, USA) at 25*C. [04451 For the aldehyde and nitroalkane-linked oligonucleotides (4b, 4c, 5b, 6a, 17, 24, and 26, Figures 46 and 47) the NHS esters were generated by combining the appropriate 30 carboxylic acid (900 mM in DMF) with equal volumes of dicyclohexylcarbodiimide (900 mM in DMF) and NHS (900 mM in DMF) for 90 minutes. Phosphorane-linked oligonucleotides (2b WO 2004/016767 PCT/US2003/025984 -141 and.20, Figures 46 and 47) were prepared by a 90 minute reaction of the appropriate amino terninated oligonucleotide with 0.1 volurhes of a 20 mg/mL DM1F solution of the NHS ester of, iodoac'etic acid (SIA, Pierce, Rockford; IL, USA) in pH 72 buffer as above, followed by' addition of 0.1 volumes of a 20 mg/mL solution of 4-diphenylphosphinobenzoic acid in DMF. 5 Thiol-linked template 16 was synthesized by reacting ethylene glycol bis(succinimidylsuccinate) (EGS, Pierce, Rockford, IL,.USA) with the appropriate oligonucleotide for 15 miniute§, followed by addition of 0.1 volumes of 300 miM 2-aminoethanethiol. Reagent 5a was synthesized using 3'-OPO 3

-(CH

2

)

3

SS(CH

2

)

3 0DMT functionalized controlled-pore glass (CPG) supjfort and reduced prior to use according to the manufacturer's protocol. 10 [0446] The 3'-biotinylated oligonucleotides were prepared using biotin-TEG' CPG (Glen Research, Sterling, Virginia, USA). Products arising from biotinylated templates were purified by mixing with 1.05 equivalents of streptavidin-linked magnetic beads (Roche), washing twice with 4 M guanidinium hydrochloride, and eluting with aqueous 10 mM Tris pH 7.6 with 1 mM biotin at 80 *C. 15 Synthesis of Linkers [0447] Linkers between DNA oligonucleotides and the functional groups in I a-6c are as follows.' lb and Ic: DNA-5'-NH2; la, 2a-2c, 3a, and 3c: DNA-5'-O(CH 2

)

2 0(CH 2

)

2 -N-H-; 5a: DNA-3-O-(CH 2

)

3 SH; 4a-4c, 5b, 5c, 6a, and 6c : DNA-3'-O-CH 2

CH(CH

2

OH)(CH

2

)

4 NH-. Oligonucleotide sequences used to generate all possible products in Figure 46 (lanes 5, 9, and 20 14), with annealing regions underlined: R-TATCTACAGAG-3' [SEQ ID NO: 106] (la-1c); R TATCTACAGAGTAGTCT-3' [SEQ ID NO: 107] (2a-2c); R TATCTACAGAGTAGTCTAATGAC-3' [SEQ ID NO: 108] (3a-3c); 5'-CAGCCTCTGTAGAT R [SEQ ID NO: 109] (4a-4c); 5'-CTCAGCCTCTGTAGAT-R [SEQ ID NO: 110] (5a-Sc); 5' GGCTCAGCCTCTGTAGAT-R [SEQ ID NO: 111] (6a-6e). Functionalized templates and 25 reagents were purified by gel filtration (Sephadex G-25) followed by reverse-phase HPLC (0.1 M triethylammonium acetate/acetonitrile gradient). Representative functionalized templates and reagents were further characterized by MALDI mass spectrometry. Reaction Conditions [0448] All reactions were performed by dissolving reagents and templates in separate 30 vessels in pure water before combining them into a solution of 50 mM aqueous TAPS buffer, pH 8.0, 250 mM NaCI at 25 4C for 16 hours with DNA-linked reactants at 60 nM (Figure 47) or at V % eU4UPCT/US2003/025984 - 142 12.5 nM (Figures 47 and 48). NaBH 3 CN, EDC, and sulfo-NHS were present when appropriate as described. Products were analyzed by denatuing polyacrylamide gel electrophoresis 1isIng ethidium-bromide staining and UV transilluinination: Differences in charge states, attached functional groups, and partial secondary structure resulted in modest variations in gel mobility 5 for different functioialized oligonucleotides of the same length (Figures 46-48). Example 8: DNA-Templated Functional Group Transformations [04491 While coupling reactions' are useful for building molecular diversity, the development of DNA-templated functional group tansformations can significantly expand the types of struefures that can be generated. DNA-templated synthesis can be used to transform. 0 functional groups by unmasking or interconverting finctionalities used in coupling reactions. By exposing or creating a reactive group within a sequence-programmed subset of a'library, DNA templated functional group interconversions permit library diversity t6 be generated by sequential unmasking (Figure 49). In Figure 49, PG1 - PG3 represent three different protecting groups, and A-F represent reactants capable of reacting with deprotected functionalities, of a 5 scaffold molecule. The sequential unmasking approach offers the major advantage of pennitting reactants that would normally lack the ability to be linked to DNA (for example, siinple alkyl halides) to contribute to library diversity by reacting with a sequence-specified subset of templates in an intermolecular, non-templated reaction mode. This advantage significantly increases the types of structures that can be generated. On the other hand, sequential unmasking O has the drawback of requiring more manipulations per "step" because previously used small molecule reactants must be removed between DNA-templated functional group unmaskings. This removal can be rapidly performed on the entire library using a simple gel filtration cartridge. DNA-Templated Deprotection 5 [0450] The first class of DNA-templated functional group transformations sequence specifically unmask amine, thiol, alcohol, carboxylate, or aldehyde groups from protected forms. In the Staudinger reaction, azides react with phosphines to yield aza-ylides (Staudinger et al. (1919) HELV. CHIM AcTA. 2: 635-646). When this reaction is performed in aqueous media, the aza-ylides undergo spontaneous hydrolysis to provide amines and phosphine oxides (Scriven et 0 al. (1988) CHEM. REV. 88: 297-368). DNA-linked aryl and alkyl phosphine reagents, when combined with azide-linked DNA templates, permit sequence-specific amine deprotection WO 2004/016767 PCT/US2003/025984 -143 (Figure 50A). DNA-linked phosphines and DNA-linked azides have both been used successfully in previous DNA-templated reactions. As an alternative DNA-tenplated amine deprotection,,the nucleophilic aromatic ipso-substitution of o-nitrobenzenesulfonamides (prepared from amines and commercially available o-nitrobenzene sulfonylchloride) can yield 5 free amines, (Figure 50B). This reaction is known to proceed efficiently in the presence of deprotonated thiophenots, so at pH > 8 the DNA-templated attack of thiophenol-linked reagents on o-nitrobenzenesulfonamide-linked templates can permit sequence-specific amine deprotection. (Fukuyama et al. (1999) SYNLETT 8: 1301-1303). [0451] Once optimized, DNA-templated amine deprotection reactions can be extended to 10 include deprotection reactions for alcohols and thiols; Kusumoto and co-workers have reported that 4-aminobutyryl esters undergo spontaneous intramolecular lactam formation to afford 2 pyrrolidinone and the liberated hydroxyl group in excellent yields (Kusumoto et atl (1986) BULL. CHEM. Soc. JPN. 59: 1296-1298). Kahne and co-workers have used this reaction effectively in aqueous inedia (Thomson et al. (1999) J. AM. CHEM. Soc. 121: 1237-1244). A DNA-templated 15 hydroxyl group deprotection is shown in Figure 50C. If lactam formation is slow, the reaction can be heated or Lewis acids can be added since sequence specificity is not required after amine deprotection. An analogous DNA-templated thiol deprotection that uses 4-azidobutyryl thioesters is shown in Figure 50C. It is contemplated that these groups will be stable to hydrolysis under a wide range of conditions. 20 [04521 Palladium-mediated deallylation can also be used in DNA-temp lated carboxylate, amine, hydroxyl, or thiol deprotections. Allyloxycarbonyl (Alloc) esters, carbonates, thiocarbonates, and carbamates are treated with DNA-linked Pd ligands such as the 2, 2' bis(diphenylphosphino)-1, 1'-binaphthyl (BINAP) reagent as shown in Figure 50D (prepared from the known BINAP-6-butanoic acid) in the presence of pM to pM concentrations of water 25 soluble Pd sources such as Na 2 PdCl4 (Bayston et aL (1998) J. ORG. CHEM. 63: 3137-3140). The DNA-linked Pd ligands increase the effective molarity of Pd at complementary templates, but not at mismatched templates, to permit the sequence-specific deprotection of carboxylate, hydroxyl, thiol, and amine groups from the corresponding Alloc esters, carbonates, thiocarbonates, and carbamates, respectively (Figure 50D) (Gen6t et aL (1994) TETRAHEDRON 30 50: 497-503). It is particularly encouraging that the rates of BINAP ligand dissociation from Pd have been measured during Pd-mediated aryl aminations and found to be much slower than the WO 2004/016767 PCTUS2003/025984 -144 rates of association and dissociation of substrate and products (Singh et a!. (2002) J. AM. CHEM. Soc. 124:,14104-1411,4). The Pd sourci andJthellNA-linked Pd ligands can be pre-in6ubated at high concentrations, and then the resulting. cbmplexes added either to complementary or mismatched templates at 60 nM concentrations,' This procedure also results in sequence-specific 5 Alloc depr'Qtection if ligand-metal dissociation-is s1ow relative to DNA annealing and}d catalyzed deallylation. [04531 Finally; transition metal salts including Sc, and Yb" are known to catalyze acetal hydrolysis to yield aldehydes (Fukuzawa et ah (2001) CHEM. LETT: 5:'430-436). Conjugating the crown ether shown in Figure 50E to oligonucleotides permits DNA-teniplated 10 aldehyde deprotections in the presence of lanth'nide triflates. These crown ether-Ln" complexes'have teen previously reported t 'catalyze aqueous aldol reactions while completely sequestering one equivalent of Ln 3 " (Kobayashi et al. (2001) ORG. LETT. 3). Aldehyde deprotection is'highly sequence-specific because the concentration of free Ln 3 should be negligible. 15 DNA-Templated Functional Group Intercon versions [0454] The second class of DNA-templated functional group transformations interconverts groups generated from or used by DNA-templated reactions. Two functional group interconversions are shown in Figure 51. Ruthenium(II) porphyrins in the presence of 2,6 disubstituted pyridine N-oxides catalyze the remarkably efficient epoxidation of a wide variety 20 of simple and electron-deficient olefins (Higuchi et al. (1989) TETRAHEDRON LETr. 30: 6545 6548; Groves et al. (1985) J. AM. CHEM. Soc. 107: 5790-5792; Zhang et a!. (2002) ORG. LETT. 4: 1911-1914; Yu et a. (2000) J. AM. CHEM. Soc. 122: 5337-5342). Single-stranded DNA is stable in the presence of aqueous tetrakis(4-carboxyphenyl) porphyrin complexed with Ru(lI), and Ru(I)-DNA conjugates have been previously reported (Hartmann et al. (1997) J. BIOL. 25 INORG. CHEM. 2: 427-432; Pascaly et a. (2002) J. AM. CHEM. Soc. 124: 9083-9092). DNA templated olefin epoxidations using DNA-linked Ru(II) porphyrin catalysts are shown in Figure 51A, which are prepared by coupling commercially available tetrakis(4-carboxyphenyl) porphyrin to amine-terminated oligonucleotides (Holmlin et al. (1999) BIoCONJUG. CHEM. 10: 1122-1130). The resulting DNA-linked porphyrin is metalated with RuJ(CO)12 as described 30 previously to afford the reagent shown in Figure 51A. This functional group interconversion WO 2004/016767 PCT/UJS2003/025984 -145 bridges several versatile reactions by permitting products of DNA-templated Wittig olefinations. aid Heck'couplings to become substrates for epoxide addition reactions. 104551, As a second functional group interconversion, lanthanide triflate-catalyzed aqueous Diels-Alder and hetero Diels-Alder cycloadditionis proceed efficiently in water, and 5 DNA-linked Lewis acid chelators such as binapthol, bis-trifylamides, or the crown ether shown in. Figure 50E permit the sequence-specific'Diels-Alder reaction between a template-linked aldehyde and a free dine in solution (Figure 51B). When Danishefsky's diene is used, this, functional group transformation provides a,p-unsaturated ketones that serve as substrates for subsequent DNA-templated conjugate addition reactions. Fully coordinated Ln 3 * complexes 10 (such as those that arise from the crown ether) have been reported to be kinetically'stable yet permit efficient catalysis through facile ligand exchange (Chappell et al. (1998) INORG. CHEM. 37: 3989-3998). Moreover, DNA-linked lanthanide complexes have been previously used as stable luminescent agents in aqueous solutions and, therefore, these complexes are compatible with the functionality present in DNA (Li et al. (1997) BIOCONJUG. CHEM. 8: 127-132). 15 Example 9: Synthesis of Exemplary Compounds and Libraries of Compounds A) Synthesis of a Polycarbanate Library 104561 This Example demonstrates a strategy for producing an amplifiable polycarbamate library. Overview 20 [04571 Of the sixteen possible dinucleotide codons used to encode the library, one is assigned a start codon function, and one is assigned to serve as a stop codon. An artificial genetic code then is created assigning each of the up to 14 remaining dinucleotides to a different monomer. For geometric reasons one monomer actually contains a dicarbamate containing two side chains. Within each monomer, the dicarbamate is attached to the corresponding 25 dinucleotide (analogous to a tRNA anticodon) through a silyl enol ether linker which liberates the native DNA and the free carbamate upon treatment with fluoride. [04581 The dinucleotide moiety exists as the activated 5'-2-methylimidazole phosphate, that has been demonstrated to serve as an excellent leaving group for template-directed oligomerization of nucleotides yet is relatively stable under neutral or basic aqueous conditions 30 (Inoue et al. (1982) J. MOL. BIoL. 162: 201; Rembold et al (1994) J. MoL. EvoL. 38: 205; Chen et al. (1985) J. MoL. BOLO 181: 271; Acevedo et aL (1987) J. MOL. BIOL. 197: 187; Inoue et al WO 2004/016767 PCT/US2003/025984 -146 (1981) J. AM. CHEM. Soc 103: 7666; Schwartz-et !. (1985) SCIENCE 228: 585). The dicarbamate moiety exists in a cyclic foirm liised through a vinyloxycarbonate linker. The vinylcarbonate group has been demonstrated to be stable in neutral or basid aqueous conditions and further has been shown to provide carbamates in very high yields upon the addition of 5 amines Olofsonet al. (1977) TETRAIEDRON LETT, 18: 1563; Olofson et-aL. (1977) TETRAHEDRON LETT 18: 1567; Olofson et a (1977)TETRAHEDRON LETT. 1,8: 1571). [0459] , When attacked by an amine from a nascent polycarbamate chain, the vinyl carbonate- linker, driven by the aromatization of m-&resol, liberates a free 'amine. This fr&e amnine subsequently serves as the nucleophile to attack the next vinyloxycarbonate, propagating the 10 polymerization of the growing carbamate chain. Such a strategy minimizes the potential for cross-reactivity and bi-directional polymenization by ensuring that only one nucleophile is present at any time during polymerization., [04601 Using the monomer described above, artificial translation of DNA'inpto a polycarbamate can be viewed as a three-stage process. In'the first stage, single stranded DNA 15 templates encoding the library are used to guide the assembly of the dinucleotide moieties of the monomers, terminating with the "stop" monomer which possesses a 3'methyl ether'instead of a 3'hydroxyl group (Figure 52). [04611 Once the nucleotides have assembled, the' "start" monomer ending in a o nitrobenzylcarbamates is photodeprotected to reveal the primary amine that initiates carbamate 20 polymerization. Polymerization proceeds in the 5' to 3' direction along the DNA backbone, with each nucleophilic attack resulting in the subsequent unmasking of a new amine nucleophile. Attack of the "stop" monomer liberates an acetamide rather than an amine, thereby terminating polymerization (Figure 53). Because the DNA at this stage exists in a stable double-stranded form, variables such as temperature and pH may be explored to optimize polymerization 25 efficiency. [04621 Following polymerization, the polycarbamate can be cleaved from the phosphate backbone of the DNA upon treatment with fluoride. Desilylation of the enol ether linker and the elimination of the phosphate driven by the resulting release of phenol provides the polycarbamate covalently linked at its carboxy terminus to its encoding single-stranded DNA 30 (Figure 54).

WO 2004/016767 PCT/US2003/025984 -147 [04631 At this stage, the polycarbamatenmg be, completely liberated from the bNA by base hydrolysis of the ester linkage. The liberated polycarbamate can be purifiedby HPLC and retested to verify that its desired, properties re intact: .The free DNA can be amplified using PCR, mutated with error-prone PCR (Cadwell & at (1,992) PCR METHODS APPL. 2: 28) or DNA 5 shuffling (Stemmer (1994) PROC. NATL.. ACAD, SC. USA 91: 10747; Stemmer (1994) NATURE 370: 389; U.S. Patent 5,811,238), and/or sequenced to reveal the primary structure of the polycarbamate polymer. Synthesis of monomer units [04641 - After the monomers are synthesized, the assembly and polymerization of th, 10 monomers on the DNA scaffold should occur spontaneously. Shikimic acid 1, available commercially, biosynthetfically (Davis (I 955) AD9. ENZYMOL, 16: 287), or by short syntheses from D-mannose (Fleet et al (1984) J. CHEM. SpC. 905; Harvey et ab (1991) TETRAHEDRON LETT. 32: 4111), serves as a convenient starting point for the monomer synthesis, The syn hydroxyl. groups are protected as the p-methoxybenzylidene, and remaining hydroxyl group as 15 the tert-butyldimetbylsilyl ether to afford 2. The carboxylate moiety of the protected shikimic acid then is completely reduced by lithium aluminum hydride (LAH) reduction, tosylation of the resulting alcohol, and further reduction With LAH to provide 3. COH C02H Me I)p-MeOC6H 4 CHO, TsOH 1) LiAlH 4 2) TBSCi, imidazole 2) TsCI, pyridine HO 1 /OH 0 1 OTas 3) LiA14 o /0TBS OH 0 0 p-MeOPh p'MeOPh 1 2 3 20 [04651 Commercially available and synthetically accessible N-protected amino acids can serve as the starting materials for the dicarbamate moiety of each monomer. Reactive side chains are protected as photolabile ethers, esters, acetals, carbamates, or thioethers. Using chemistry previously developed (Cho et al (1993) SCIENCE 261: 1303), a desired amino acid 4 is converted to the corresponding amino alcohol 5 by mixed anhydride formation with 25 isobutylchloroformate followed by reduction with sodium borohydride. The amino alcohol then is converted to the activated carbonate by treatment with p-nitrophenychloroformate to afford 6, WO 2004/016767 PCT/US2003/025984 - 148 which then is coupled to a second amino alcohol 7 to provide, following hydroxyl group silylation and FMOC deprotection, carbamate 8. Q IocoaN H NHFMOC, 2 o 4 NHFMOc pyridine p-NO2Ph0 0NHFMOC HO - 7 -*- IpHO p-NO 2 PbO 0 4 R, 5 RI 1)

N

2 TESO NH I NH 0 2) TESCi, inid. 3) piperidine RI [0466] Coupling of carbamate 8 onto the shikimic acid-derived linker'proceeds as follows. The allylic hydroxyl group of 3 is deprotected with tetra-butylammonium fluoride (TBAF), treated with triflic anhydride to form the secondary triflate, then displaced with aminocarbamate 8 to afford 9. Presence of the vinylic methyl group in 3 should assist in 3 minimizing the amount of undesired product resulting from S 2 ' addition (Magid (1980) TETRAHEDRON 36: 1901). Michael additions of deprotonated carbamates to c;#-unsaturated esters have been well documented (Collado et al (1994) TETRAHEDRON LETT. 35: 8037; Hirama et al (1985) J. AM. CHEiM. Soc. 107: 1797; Nagasaka et al (1989) HETEROCYCLES 29:155; Shishido et al (1987) J. CHEM. Soc. 993; Hirama et at. (1989) HETEROCYCLES 28: 1229). By 15 analogy, the secondary amine is protected as the o-nitrobenzyl carbamate (NBOC), and the resulting compound is deprotonated at the carbamiate nitrogen. This deprotonation can typically be performed with either sodium hydride or potassium tert-butyloxide (Collado et al. (1994) supra; Hirama et al. (1985) supra; Nagasaka et aL (1989) supra; Shishido et al. (1987) supra; Hirama et at (1989) supra), although other bases may be utilized to minimize deprotonation of 20 the nitrobenzylic protons. Additions of the deprotonated carbamate to o-unsaturated ketone 10, followed by trapping of the resulting enolate with tert-butyldimethyl silyl chloride (TBSCl), should afford silyl enol ether 11. The previously found stereoselectivity of conjugate additions to 5-substituted enones such as 10 (House etaL (1968) J. ORG. CHEM, 33: 949; Still etaL (1981) TETRAHEDRON 37: 3981) suggests that 11 should be formed preferentially over its diastereomer. 25 Ketone 10, the precursor to the fluoride-cleavable carbamate-phosphate linker, may be WO 2004/016767 PCT/US2003/025984 149 synthesized from 2by one pot decarboxylation (Ba4on et aL (1985) TETRA'HEDRON 41: 3901) followed by treatment With tetrabutylaminonium fluoride (TBAF), Sworn oxidatioh of th6 resulting alcohol to ford 12, deprotection -with 2, 3-dic1j1dro-5, 6-dicyano1, 4-benzoquinone (DDQ)),selective nitrob6nzyl ether fonnation of ihe-less-iindered alcohol, and reduction of the 5 a-hydroxyl' group with samarium iodide (MoIender (1.994) ORGANIC REACTIONS 46: 211). - OCETS 2)Tf20N.P'C H OTES \_ 0 R 2 MeOPh 3 p -M eOPhBN CO,'" 1),DQ 2) Th, 3 SnH O O NHOTOTS 2) 0Ou N N\ 4) DMS', 4)TBSCL p-MDo o DN NH ciopc N 30 0 N S p-MeoPh I 104671 The p-methoxybenzylidiene group of 11 is transformed into the a-hydroxy p 10 methoxybenzyl (PMB) ether using sodium cyanoborohydride and trimethylsilyl chloride (TMSCI) (Johansson et aL (1984) J. CHEM. Soc. 2371) and the TES group deprotected with 2% HF (conditions that should not affect the TBS ether (Boschelli et al (1985) TETRAHEDRON LETT. 26: 5239)) to provide 13. The PMB group, following precedent (Johansson et aL (1984) J. CHEM. Soc. 2371; Sutherlin et aL (1993) TETRAHEDRON LETT. 34: 4897), should remain on the 15 more hindered secondary alcohol. The two free hydroxyl groups may be macrocyclized by very slow addition of 13 to a solution ofp-nitrophenyl chlorofornate (or another phosgene analog), providing 14. The PMB ether is deprotected, and the resulting alcohol is converted into a triflate and eliminated under kinetic conditions with a sterically hindered base to afford WO 2004/016767 PCT/US2003/025984 L- 150 vinyloxycarbonate 15. Photodeprotection of the nitrobenzyl either and nitrobenzyl carbamate yields alcohol 16.

C

6

-

4 OMe o .OEs OH OH I)NaBpMTN, Q\PMB2 O 2 TMS2 NN O N OONB -- 1-

-

0 N ' ONB O oNB O ONB 13 OTBS OTBS R R-R2 ONB N .) N) 4) h ,CY4j~ OPMB SOTBS

-

O OTBS i5: R-NB, R CONB /165: R41, R-If R1 R1 ONB R 5 (04681 The monomer synthesis is completed by the sequential coupling of three components. Chlorodiisopropylaminophosphine 17 is synthesized by the reaction of PC1 3 with diisopropylamine (King et aL (1984) J. ORG. CHEM. 49: 1784). Resin-bound (or 3'-o nitrobenzylether protected) nucleoside 18 is coupled to 17 to afford phosphoramidite 19. 10 Subsequent coupling of 19 with the nucleoside 20 (Inoue et at (1981) J AM. CHEM. Soc. 103: 7666) provides 21. Alcohol 16 then is reacted with 21 to yield, after careful oxidation using m chloroperbenzioc acid (MCPBA) or 12 followed by cleavage from the resin (or photo deprotection), the completed monomer 22. This strategy of sequential coupling of 17 with alcohols has been successfully used to generate phosphates bearing three different alkoxy 15 substituents in excellent yields (Bannwarth et aL (1987) HELV. CHIM. AcTA 70: 175).

WO 2004/016767 PCT/US200I3/025984 -151 P4j C HO -Pr i Pr JPr 17 Me M NN-P- 0 S 0 0 . . L I Me I TBSO 10 / Pi 0 0 -// t "111R, 104691 The uni que start and stop monomers used to initiate and terminate carbaiate polymerization may be synthesized by simple modification of the above scheme. 5 B) Macrocyclic 17'maraynide Library [04701 This Example demonstrates that DNA teruplated-synthesis can be used to create a library of small molecules, In particular, it has been possible to create a DNA-templated macrocyclic fumaramide library as shown in Figure 55. 10471] The library synthesis schemeff employs -robust DNA-temnplated amnine acylation 10 and intramolecular Wittig olefination reactions to generate diverse and partially rigid macrocyclic tinnararnides. The flimaramide group is stable to neutral solutions but is sufficiently electrophilic to covalently capture nucleophiles when presented at elevated effective molarities. Nucleophilic side chains found in target protein active, sites may, therefore, be covalently trapped by the fiunaramide functionality. The key steps in the library synthesis are (i) 15 I)NA-ttfplatecl amnine acylation using ffie sul fone linker, (Ui) D)NA-ternplated amnine acyl ation using the diol linker, (iii), DNA-temnplated amnine acylation using a phosphorane linker, and (iv) intramolecular Wittig olefinaton to afford macrocyclie furnaramides linked to their corresponding DNA templates (Figure 55).

WO 2004/016767 PCT/US2003/025984 -152 [04721 Macrocyclization is potentially the most challenging step of the library synthesis. To test this, seven model step 3 substrates were prepared to -validate the third DNA-templated step aid the subsequent macrocyclizationl (Figure 56). Ech substrate contained a- variety of R and R 2 groups of varying steric hindrances, stereochemistries, and backbone chain lengths. The 5 model substrates were each mixed with one of four biotinylated DNA-linked reagents containing both a carboxylic acid and a phosphorane under DNA-templated amine acylation conditions. To evaluate both aide bond formation and Wittig macrocyclization, a two-stage purification strategy was implemented. The-ten products of the DNA-templated amine acylation (Figure 56 and step 3 in Figure 55) were purified away from unreacted templates by capture with 10 streptavidin-linked magnetic beads. The captured intermediates then were treated with pH 8.0 buffer to induce Wittig olefination-mediated macrocyclization. Macrocyclizatidn created the fumaramide products (lacking the biotinylated reagent oligonucleotide) to self-elute from the magnetic beads. In every case, amine acylation and macrocyclization proceeded efficiently (Figure.56) despite the wide range of steric, stereochemical, and backbone diversity in the 15 intermediates. Control reactions at pH _ 6 (too low to form the phosphorane), or at pH 8.0 but lacking the aldehyde group, failed to elute any product. In summary, the DNA-templated amine acylation-Wittig macrocyclization sequence is a highly efficient route to produce desired macrocyclic funaramides. [0473] After validating the macrocylization step, a DNA-templated macrocyclic 20 fumaramide library was synthesized. The pilot library was restricted to 83 macrocyclic fumaramides containing 4 x 4 x 5 =80 macrocycles plus three macrocycles containing either an aryl sulfonamide, a desthiobiotin group, or both groups as positive controls for binding to carbonic anhydrase or avidin. Reagent oligonucleotides consisted of the six-base codons flanked by two constant bases on either side conjugated at their 3' ends to aminoacyl donors through the 25 sulfone, diol, or phosphorane linker as previously reported. Multi-jg quantities of each of the 19 DNA-linked amine acylation reagents shown in Figure 57 were created in a single day starting from commercially available free amino acids, linker precursors, and reagent oligonucleotides as described previously. The building blocks were chosen to sample structural and functional group diversity and include (L) and (D) a-amino acids, a,a'-disubstituted amino acids, and p 30 amino acids bearing alkyl, alkenyl, aryl, polar, heterocyclic, negatively charged, and positively charged side chains (Figure 57). Each of the 19 reagents was successfully tested in single template reactions and generated product with < 30% variance in efficiency. All 19 reagents WO 2004/016767 PCT/US2003/025984 -153 reacted with high sequence-specificity, generating no significant product with mismatched teiplates even when five equivalents of reagent were used. [04741 The macrocyclic fumaramide-encoding template library was prepared from modular coding region cassettes in a single solution (Figure 58). Oligonucleotides representing 5 all reagent annealing regions were combined together with T4 DNA ligase in a single solution. Due to the sequence design of the oligonucleotide termini, the desired assembled template library is the only possible product when the ligation is complete. Excellent yields of the desired template library resulted from a 4 hour ligation reaction. Following ligation, T7 exonuclease was added to degrade the non-coding template strand'(the desired coding strand is protected by 10 its non-natural 5-aminoethylene glycol linker). This -procedure provided 20 nmol of the 5' funetionalized single-stranded template library in 6 hours. The constant I 0-base primer binding regions at the ends of each template were sufficient to permit PCR amplification of as few as 1,000 molecules (1021 mol) of template from this assembled material. Three positive control templates were added to produce a library containing 83 templates which were then combined 15 with 3.0 equivalents of five step 1 reagents to produce the first library synthesis step. Products were purified as described above, then subjected to the second DNA-templated library synthesis step with five new reagents complementing the step 2 coding regions. The efficiency of both DNA-templated pilot library steps was judged to exceed 70% by denaturing gel electrophoresis and densitometry. 20 10475] As a model for the deprotection prior to step 3, the Pd-mediated deprotection of DNA-linked Alloc carbamates was executed with excellent efficiency as judged by the liberation of -1 equivalent of free amine groups. The products from each library synthesis step were analyzed by mass spectrometry. In the hope of eliminating the deprotection step, the necessity of protecting and deprotecting the side chain amine in the starting material was tested because the 25 lower pKa of the a-amine may permit selective reaction of the a-amine at a pi that ensures protonation of the side chain amine. It was found that the a-amine group indeed could be selectively and efficiently acylated in a DNA-templated reaction in the presence of unprotected side-chain amine at pH 6.0. This may eliminate the need for a deprotection step following the second DNA-templated amide formation in step 2. 30 [0476] Several model substrates then were synthesized to validate the third DNA templated step and the subsequent macrocyclization. Each model substrate consisted of a WO 2004/016767 PCT/US2003/025984 -154 template-linked intermediate containing a free amine group and a diol linker separated by varying nurnbers of bonds to simulate groups Qf differing, sizes during library synthesis. The model substrates were each mixed with one of several biotinylated DNA-di'ked reagents containing both a carboxylic acid and a phosph6rane under DNA-templated aide formation 5 conditions (pH 6.0, 20 mM EDC, 15 mM sulfo-NHS). DNA-templated amide formation proceeded in >60% yields and products werelcaptured with avidin-linked magnetic beads. Bead bound product was treated with 10 mM NaIO4 at pH 8.5to effect diol cleavage. The resulting aldehydd grdup reacted with the phosphorane in a spontaneous Wittig olefination reaction to furnish a cyclic fumaramide, free from the biotin group, that self-elutes from the avidin-linked 10 beads (Figure 59). Importantly, all of the model substrates under went macrocyclization in >60% yield, suggesting that this reaction is tolerant of a variety of substrate geometries. Control reactions confirmed that fumaramide formation was dependent on (i) periodate cleavage, (ii) the presence of the phosphorane group, and (iii) sucbessful DNA-templated amide formation (required for capture onto avidin-linked beads). 15 C) PNA Polymer Library Formation [04771 Despite significant successes, the generality and sequence-specificity of template directed polymerization is still largely unexplored. For example, the efficient and sequence specific templated polymerization of easily functionalized synthetic monomers lacking a ribose backbone has not been reported. Such a system would raise the possibility of evolving polymers 20 comprised of these synthetic monomers through iterated cycles of translation (polymerization), selection, and amplification presently available only to DNA, RNA, and proteins. [04781 The minimal requirements of a system for synthetic polymer evolution are: (i) distance-dependent nucleic acid-templated monomer coupling reactions to ensure that oligomerization proceeds exclusively between adjacently annealed monomers; (ii) efficient 25 nucleic acid-templated oligomerization to provide sufficient yields of full-length products for in vitro selections; (iii) stable linkage of each synthetic polymer to its encoding template to ensure the survival of the appropriate template during polymer selection; and (iv) a readily functionalized synthetic monomer backbone to introduce tailor made functionality into the polymer. 30 [0479] In order to test the feasibility of producing polymers by DNA templated synthesis, DNA-templated amine acylation, Wittig olefination, reductive amination, and olefin metathesis WO 2004/016767 PCTIJS2003/025984 -155 reactions were tested for their ability to translate DNA sequences into functionalized peptide nucleic aeid (PNA) polymers. The proposed PNA monomers are stable and can be easily synthdsized from commercially availabldao-amino acids c'dritaining a wide variety of functional groups (Haaima et at. (1996) ANGEW. CHEM. INT. ED. ENGL. 35: 1939-1942; Puschl et al. (1998) 5 TETRAHEDRON LETr. 39: 4707). PNAs containing finctionalized side chains are known to retain their ability to hybridize to DNA sequence-specifically (Haaima et al. (1996) supra; Puschl et at. (1998) supra) 19480] In the first strategy, PNA serves as the backbone of the functional polymer and displays the functional groups of each monomer. In another strategy, the DNA-templated PNA 10 polymerizations organize reactive functional groups, enabling a second polymerization reaction between these functional groups (for example, an olefin metathesis or Wittig olefination reaction) to form the synthetic polymer backbone of interest. 104811 In both strategies templates consist of 5-functionalized, single-stranded DNA libraries 50-200 bases long that contain a central region of variable bases. These templates are 15 made by standard solid-phase oligonucleotide synthesis combined with enzyme-catalyzed ligation for longer templates. Monomer structures are chosen to provide chemical functionalities including (i) Bronsted acidic and basic groups, (ii) nucleophilic and electrophilic groups, (iii) conjugated olefins suitable for post-PNA polymerization metathesis, and (iv) metal-binding groups capable of forming complexes with chemically potent transition metals. Representative 20 monomer structures containing these functionalities are shown in Figure 60. The DNA bases encoding each monomer (the "genetic code" of these polymers) are chosen from the examples shown in Table 10 to preclude the possibility of out-of-frame annealing. These genetic codes should prevent undesired frameshifted DNA-templated polymer translation. [0482] Libraries of 5'-functionalized hairpin DNA templates containing up to 105 25 different sequences are combined with sets of monomers under conditions that optimize the efficiency and sequence fidelity of each DNA-templated polymerization. Synthetic polymer strands then are de-annealed from their DNA templates by denaturation, and the 3' DNA hairpin primer extended using DNA polymerase to generate hairpin DNA templates linked to now liberated single-stranded synthetic polymers (Figure 61). Libraries are characterized by gel 30 electrophoresis and MALDI mass spectrometry, and individual representative library members are also characterized from single template reactions to confirm expected reaction efficiencies.

WO 2004/016767 PCT/US2003/025984 156 [0483] Once the libraries of DNA-linkedPNAs are characterized, they can be subjected to three types of in vitro selections for: (i) fohding; (ii) target binding, or (iii) catalysis. Prior to selection, polymers-with anticipated m6tal binding ability are incubated with one or more water compatible metal sources. Selections forfolding' are performed using the gel electrophoresis 5 selection described in' Example 10. Polymerscapable of folding in the presence, but not in the absence, of metals serve as especially attractive starting points for the next two types, of selections. [04841 Selections for target binding can be conducted by incubating tie solution-phase polymer library with either immobilized target or with biotinylated target followed by 10 streptavidin-linked beads. Non-binders are removed by washing, and polymers with desired binding properties are eluied by chemical deriaturation or-by adding excess authentic free ligand. To complete one cycle of functionalized PNA evQlutioif, the DNA templates corresponding to the desired' PNA library members are amplified by PCR using one primer containing the 5' functionalized hairpin primer and a biotinylated second primer, optionally diversified by error 15 prone PCR (Caldwell et at. (1992) PCR METHODs APPLic. 2: 28-33), and then denatured into single stranded DNA and washed with streptavidin beads to remove the non-coding template strand. The resulting pool of selected single-stranded, 5-functionalized DNA completes the evolution cycle and enters subsequent rounds of DNA-templated translation, selection, diversification, and amplification. 20 [04851 Selection for synthetic polymers that catalyze bond-forming or bond-cleaving reactions can also be performed. To select for bond-forining catalysts (for example, hetero Diels-Alder, Heck coupling, aldol reaction, or olefin metathesis catalysts), functionalized PNA library members are covalently linked to one substrate through their 5' hairpin termini. The other substrate of the reaction is synthesized as a derivative linked to biotin. When dilute 25 solutions of library-substrate conjugate are reacted with the substrate-biotin conjugate, those library members that catalyze bond formation induce self-biotinylation. Active bond forming catalysts then are separated from inactive library members by capturing the former with immobilized streptavidin. In an analogous manner, functionalized PNAs that catalyze bond cleavage reactions such as retro-aldol reactions, amide hydrolysis, elimination reactions, or 30 olefin dihydroxylation followed by sodium periodate cleavage can also be selected. In this case, library members are linked to biotinylated substrates such that the bond breakage reaction causes WO 2004/016767 PCT/US2003/025984 -157 the disconnection of the biotin moiety from the library members. Active catalysts.self-elute from. streptavidin-linked beads while inactive catalysts remain bound. Validation of PNA Polymer Library Formation [0486} Peptide nucleic acids (PNAs) are attractive 'candidates for synthetic polymer 5 evolution because of their known ability to bind DNA sequence-specifically, and their simple preparation from synthetically accessible amino acids, Previous efforts to oligomerize PNAs-on DNA or RNA templates have used amine acylation as the coupling reaction and proceeded with modesf efficiency and sequence specificity (Bohler al. (1995) NATURE 376: 578-581; Schmidt et al. (1997) NUC. ACIDS RES. 25: 4792-4796). 10 When five PNA tetramers were combined using a variety of aqueous amipe acylation conditions in the presence of DNA templates containing complementary 20-base annealing regions, only modest formation (<20% yield) of full-length PNAs, representing five successive coupling reactions, were observed. Even more problematic, however, was the formation of higher molecular weight products indepe ndent of the position of a mismatched 4base'annealing 15 region in the template. These observations indicate that PNAs are able to couple using amine acylation chemistry even when not adjacently annealed, leading to an unpredictable mixture of products. 10487] It was contemplated that the distance independence previously observed in DNA templated amine acylation reactions was the origin of the poor regiospecificity of amine 20 acylation-mediated PNA couplings. This Example shows that it is possible to overcome this problem by replacing the distance independent amine acylation reaction with a distance dependent DNA-templated reaction, such as a reductive amination reaction. [0488] In order to test this, a thymine-containing PNA monomer amino aldehyde was synthesized and coupled to threonine-linked resin following the method of Ede and Bray (Ede et 25 al. (1997) TETRAHEDRON LETTERS 38, 7119-7122). Standard FMOC peptide synthesis was used to extend the peptide by three PNA monomers (final sequence: NH 2 -gact-CHO), and aqueous acidic cleavage from the resin yielded the desired tetrameric peptide aldehyde I (Figure 62). [04891 A DNA template containing a 5'-amine-terminated hairpin and five successive 30 repeats of the "codon" complementary to 1 (5'-AGTC-3') was combined with 8 [M 1 in aqueous p1 8.5 buffer. The reactants were annealed (95 0 C to 25oC) and NaCNBH 3 was added WO 2004/016767 PCT/US2003/025984 15 -158 to 80 mM. The reactions were quenched by buffer exchange with a Sephadex column, and subjected to tion (95C for 10 tinuies in 50% formamide) and 15% denaturing'PAGE: sujetdenauralonn50 frmm In Figure 62, lanes i and 2 show that the starting template was almost entiiely consumed, and the higher molecular weight product was formed in >90% yield. Gel purification of the product 5 following removal of the DNA template-with DNase I and MALDI-TOF mass spectronpetry confirmed full-length pentamer of thegact PNA aldehyde. This result indicates that DNA templated reductive animation can mediate fhi&Iighly efficient oligomerization of PNA aldehydek [0490) In order to examine the regio- and' sequenc6-specificity of this reaction, the 10 oligomerization reactions'were repeated using a variety of template sequences. When a mismatched DNA template codon (5'-ATdC--3') was introduced at the second, third, fourth, or fifth 4-base coding region (i.e., the codon) of etemplAte, highly efficient formation of products corresponding to the coupling of exactly one, two, three, or four copies of 1, respectively, was observed (see, Figure 62, lanes 4-14). When the mismatched codon was placed at only the first 15 coding position, or at all five coding positions, no product formation was observed (see, Figure 62, lanes 3 and 15). The termination of oligomerization at the first mismatched codon in every case indicates that the DNA-templated PNA aldehyde coupling requires functional group adjacency (i.e., is highly distance dependent), and, therefore, is ideally suited for templated polymerization. 20 [04911 The sequence specificity of this system was probed by performing oligomerization experiments using DNA templates containing eight different mismatched codons (ATTC, ATGC, ATCC, AGGC, AGCC, ACTC, ACGC, or ACCC) in the third coding region. Even though four of these codons differ from the matched sequence (ATGC) in only one base, in each case only two copies of I were coupled to the template (see Figure 62, lanes 5-12). This 25 high degree of sequence specificity raises the possibility that libraries of different DNA sequences may be faithfully translated into libraries of corresponding polymers using this system, analogous to DNA-templated small molecule synthesis. [0492] It is contemplated that synthetic polymers with desired properties (e.g., binding or catalytic properties) may require lengths beyond those previously achieved efficiently using 30 nucleic acid-templated synthesis. In order to test the ability of the above system to generate longer polymers in an efficient and sequence-specific manner, DNA templates were translated WO 2004/016767 PCT/US2003/025984 -159 with 40-base coding regions encoding ten repeats of the, above matched or mismatched codon into corresponding PNA aldehyde polydtersPolyinerizations were carried out as'in Figure 62' except that the.PNA"peptide aldehyde 'concentration was 16 pM and the reaction time with NaCNBH3 was 15 minutes. The results of these' experiments are shown in Figure 63, where the 5 lanes alternate betwen -template (with.mismatch at indicated position) and reactions (ternplate plus the gact monomer). As Figure 63 illustrates, both denaturing PAGE anO MALDI-TOF' mass spectrometry revealed a single predQmifni4t product corresponding to the polymerizatidn of a full length 40-mer PNA after 15 miriutes. Introducing a inismatched codon-in the first,,third, fifth, seventh, or ninth coding positions on the terpplate again resulted in truncation,(Figure.63, 10 lanes 4, 6, 8, lO, and 12, respectively). This efficient translation of DNA sequences into 40 PNA bases (10 couplings) provides a polymer of length similar to DNA and RNA oligonucleotides with binding or catalytic properties, but mad erntirely of synthetic building blocks. [04931 A challenging requirement of creating libraries of sequende-defined synthetic polymers in this manner is maintaining sequente specificity in the presence of multiple 15 monomers of closely related sequence. In order to study the specificity of DNA-templated polymerization using multiple PNA building blocks in a single solution, nine PNA aldehyde tetramers of the sequence NH 2 -gvvt--CHO (v = g, a, or c) were synthesized. In addition, nine DNA templates containing one of nine codons complementary to gvvt at codon 5, and containing AGTC at the other nine positions were prepared. Reaction conditions were identical to those 20 from Figure 63, except that the reaction time with NaCNBH 3 was further shortened to 5 minutes and incubation was carried out at 37"C. The first two lanes of each panel in Figure 64 show a positive control polymerization. Each additional set of four lanes corresponds to: (i) 20 pmol template, (ii) reaction with 14.4 pM gact, (iii) reaction with 14.4 pM gact plus 1.6 pM PNA aldehyde complementary to the highlighted codon, and (iv) reaction with 14.4 RM gact plus 0.2 25 yM of each PNA aldehyde of the sequence gvvt except the PNA complementary to the highlighted codon. As expected, each of the nine templates was translated into a single predominant truncated product corresponding to the incorporation of four copies of I when 1 was the only PNA building block included in the reaction (37 'C, 5 min) (see, Figure 64). Full length product was efficiently generated for all nine templates, however, when the PNA 30 aldehyde complementary to the fifth coding sequence was included in addition to 1. When all PNA aldehyde tetramers were included in the reaction except the PNA complementary to the fifth coding region, only the truncated product was efficiently generated (see, Figure 64).

WO 2004/016767 PCT/US2003/025984 -160 [04941 Taken together, these experiments reveal that DNA-templated PNA aldehyde polymerizations maintain sequence specificity even when a mixture of different PNA building blocks'are present in a single solution. D) Evolving Plastics 5 {0495] In yet another embodiment,,a nucleic acid (e.g., DNA, RNA, derivative thereof) is attached to a polymerization catalyst. Since 'nucleic acids can fold into complex structures, the nucleic acid can be used to direct and/or affect the pdlymerization of a growing polymer chain. For example, the nucleic acid may influence the selection of monomer units to be polymerized as well as how the polymerization reaction takes place (e.g., stereochemistry, tacticity, activity). 10 The synthesized polymers may be selected for specific properties such molecular, eight, density, hydrophobicity, tacticity, stereoselectivity, etc., and the nucleic acid which formed an integral part of the catalyst which directed its synthesis may be amplified and evolved (Figure 65A). Iterated cycles of ligand diversification, selection, and amplification allow for the true evolution of catalysts and polymers towards desired properties. 15 104961 By way of example, a library of DNA molecules is attached to Grubbs' ruthenium-based ring opening metathesis polymerization (ROMP) catalyst through a dihydroimidazole ligand (Scholl et al. (1999) ORG. LETT. 1(6): 953) creating a large, diverse pool of potential catalytic molecules, each unique by nature of the functionalized ligand (see, Figure 65B). Functionalizing the catalyst with a relatively large DNA-dehydroimidazole (DNA 20 DHI) ligand can alter the activity of the catalyst. Each DNA molecule has the potential to fold into a unique stereoelectronic shape which potentially has different selectivities and/or activities in the polymerization reaction (Figure 66). Therefore, the library of DNA ligands can be "translated" into a library of plastics upon the addition of various monomers. In certain embodiments, DNA-DHI ligands capable of covalently inserting themselves into the growing 25 polymer, thus creating a polymer tagged with the DNA that encoded its creation, are used. Using the synthetic scheme shown in Figure 65A, dehydroimidazole (DHI) ligands are produced containing two chemical handles, one used to attach the DNA to the ligand, the other used to attach a pedant olefin to the DHI backbone. Rates of metathesis are known to vary widely based upon olefin substitution as well as the identity of the catalyst. Through alteration of these 30 variable, the rate of pendant olefin incorporation can be modulated such that kpendant olefin metathesis << kRMp, thereby, allowing polymers of moderate to high molecular weights to be formed WO 2004/016767 PCT/US2003/025984 -161 before insertion of the DNA tag and corresponding polymer termination. Vinylic ethers are commonly used in RQMP to functinalize the polymer termini (Gordon et at (2000) CHEM. BIoL. 7: 9-16), as well as produce polymers'of decreased molecular weight 10497] A polymer from the library is subsequently selected based on a desired property 5 by electrophoresis, gel filtration, centrifugal sedinientation, partitioning into solvents of different hydrophobicities, etc. Amplification and diversification of the coding nucleic acid via techniques spch as error-prone PCR or DNA shufflirig followed by attachment to a DHI. backbone'will allow for production of another pool of potential ROMP citalysts enriched in the selected activity (Figure 66). This method provides a new approach to generating polymeric, 10 materials and the catalysts that create them. Example 10: Development of Catalysts by-Templated Synthesis 104981 An altimative approach to translating DNA into non-natural evolvable polymers takes advantage of the ability of some DNA pplymerases to accept certain modified nucleotide triphosphate substrates (Perrin et al. (2001) J. AM. CHEM. Soc. 123: 1556;' Perrin et a (1999) 15 NUCLEOSIDES NUCLEOTIDES 18: 377-91; Gourlain et al (2001) NUCLEiC ACIDS RES. 29: I898 1905; Lee et al. (2001) NUCLEIC ACIDS RES. 29: 1565-73; Sakthievel et al. (1998) ANGEW. CHEM. INT. ED. 37: 2872-2875). Several deoxyribonucleotides and ribonucleotides bearing modifications to groups that do not participate in Watson-Crick hydrogen bonding are known to be inserted with high sequence fidelity opposite natural DNA templates. Importantly, single 20 stranded DNA containing modified nucleotides can serve as efficient templates for the DNA polymerase-catalyzed incorporation of natural or modified mononucleotides. 104991 The functionalized nucleotides incorporated by DNA polymerases to date are shown in Figure 67. In one of the earliest examples of modified nucleotide incorporation by DNA polymerase, Toole and co-workers reported the acceptance of 5-(1-pentynyl)-deoxyuridine 25 1 by Vent DNA polymerase under PCR conditions (Latham et al (1994) NUCLEIC ACIDS RES. 22: 2817-22). Several additional 5-functionalized deoxyuridines (2-7) derivatives were subsequently found to be accepted by thermostable DNA polymerases suitable for PCR (Sakthievel et al. (1998) supra). The first functionalized purine accepted by DNA polymerase, deoxyadenosine analog 8, was incorporated into DNA by T7 DNA polymerase together with 30 deoxyuridine analog 7 (Perrin et al. (1999) NUCLEOSIDES NUCLEOTIDES 18: 377-91). DNA libraries containing both 7 and 8 were successfully selected for metal-independent RNA cleaving WO 2004/016767 PCT/US2003/025984 -162 activity (Perrin et at. (2001) J. Am. Chem. Soc. 123:,1556-63). Williams and co-wQrkers recently tested several deoxyuridine derivatives for acceptance by Taq DNA polymerases and concluded that acceptance is greatest wbi using C5-modfied uridines bearing rigid alkyne or trans-alkene groups such as 9 and 10 (Lee et at. (2001) NUCLEIC ACIDs REs. 29: 1565-73). A 5 similar study (Gourlain et al. (2001) NUCLEIC ACIDs REs. 29: 1898-1905) on C7-functionalized 7-deaza-deoxyadenosines revealed acceptance by Taq DNA polymerase of 7-aminopropyl- (11), cis-7-aminiopropenyl- (12), and 7-aminopropynyl-7-deazadeoxyadefnosine (13). [05001 With simple general acid pd general base functionality, chiral metal centers would expand considerably the chemical scope of nucleic acids. Functionality aimed at binding 10 chemically potent metal centers has yet to been incorporated into nucleic acid polyniers. Natural DNA has demonstrated the ability to fold in complex three-dimensional structures capable of stereospecifically binding target molecules (Lin et al. (1997) CHEM. BIOL. 4: 817-32; Lin et al (1998) CHEM. BIOL. 5: 555-72; Schultze bt al. (1994) J. MOL. BIOL. 235: 1532-47) or catalyzing phosphodiester bond manipulation (Santoro et aL (1997) PRoc. NATE. ACAD. Sci. USA 94: 15 4262-6; Breaker et a. (1995) CHEM. BIoL. 2: 655-60; Li et al. (2000) BIOCHEMISTRY 39: 3106 14; Li et al. (1999) PRoc. NATL. AcAD. Sci. USA 96: 2746-5 1), DNA depurination (Sheppard et al. (2000) PROc. NATL. ACAD. Sc. USA 97: 7802-7807) and porphyrin metallation (Li et al. (1997) BIoCHEMISTRY 36: 5589-99; Li et al (1996) NAT. STRUCT. BIoL. 3: 743-7). Non-natural nucleic acids augmented with the ability to bind chemically potent, water-compatible metals 20 such Cu, La, Ni, Pd, Rh, Ru, or Se may possess greatly expanded catalytic properties. For example, a Pd-binding oligonucleotide folded into a well-defined structure may possess the ability to catalyze Pd-mediated coupling reactions with a high degree of regiospecificity or stereospecificity. Similarly, non-natural nucleic acids that form chiral Sc binding sites may serve as enantioselective cycloaddition or aldol addition catalysts. The ability of DNA polymerases to 25 translate DNA sequences into these non-natural polymers coupled with in vitro selections for catalytic activities would therefore permit the direct evolution of desired catalysts from random libraries. [0501] Evolving catalysts in this approach addresses the difficulty of rationally designing catalytic active sites with specific chemical properties that has inspired recent combinatorial 30 approaches (Kuntz et al. (1999) CURR. OPIN. CHEM. BIOL. 3: 313-319; Francis et at (1998) CURR. OPIN. CHEM. BIOL. 2: 422-8) to organometallic catalyst discovery. For example, WO 2004/016767 PCT/US2003/025984 -163 Hoveyda and co-workers identified Ti-based enantibselective epoxidation catalysts by serial screening of' eptide ligands (Shimiiu etl!.(1997). ANGEW. CHEM. INT. ED. 36). Serial screening was also used by Jacobsen aid co-$vorkers to identify peptide liginds that form enantioselective epoxidation catalysts when coniplexed. With metal catiois (Francis et al. (1999) 5 ANGEW. CfLEM. INT. ED-. ENGL. 38: 937-941).' Recently,. a peptide library containing phpsphine side chains was screened for the ability to catadlyze malonate ester addition tocyclopentenyl acetate in the presence of Pd (Gilbertson et al:000)VAM. CHEM. Soc. 122: 6522-6523). [0502] The current approach differs fundamentaJly from previous bornbinatorial catalyst discovery efforts in that it permits catalysts with desired properties to spontaneously eme-ge , 10 from one pot, solution-phase libraries after evolutionary cycles of diversification, amplification, translation, and selection. This strategy allows up to 105 different catalysts to be-genierated and selected for desired properties in a single experiment. The compatibility of this approach with one-pot in vitrd selections allows the direct selection for reaction catalysis rather than screening for a phenomenon associated with catalysis such as metal binding or heat generation. In 15 addition, properties difficult to screen rapidly such as substrate stereospecificity or metal selectivity can be directly selected using approaches disclosed herein. [05031 Key intermediates for a number of C5-functionalized uridine analogs and C7 functionalized 7-deazaadenosine analogs have bebn synthesized for incorporation into non natural DNA polymers. In addition, the synthesis of six CS-functionalized adenosine analogs as 20 deoxyribonucleotide triphosphates has been completed. Synthesis of Metal-Binding Nucleotides [0504] A strategy for synthesizing metal-binding uridine and 7-deazaadenosine analogs is shown in Figure 68. Both routes end with amide bond formation between NHS esters of metal-binding functional groups and amino modified deoxyribonucleotide triphosphates (7 and 25 13). Analogs 7 and 13 as well as acetylated derivatives of 7 have been previously shown to be tolerated by DNA polymerases, including thermostable DNA polymerases suitable for PCR (Perrin et al (2001) supra; Perrin et al, (1999) supra; Latham et al. (1994) NUCLEIC ACIDS 12vIPRE 22: 2817-22; Gourlain et a. (2001) Nucleic Acids Res. 29: 1898-1905; Lee et at. (2001) NUCLEIc ACIDS RES. 29: 1565-73; Sakthivel et aL (1998) ANGEW. CHEM. INT. ED. ENOL. 37: 30 2872-2875). This approach allows a wide variety of metal-binding ligands to be rapidly incorporated into either nucleotide analog. Amino modified deoxy-ribonucleotide triphosphate 7 WO 2004/01.6767 PCT/US2003/025984 -164 has been synthesized using a previously reported route (Sakthivel et al. (1998) supra). As illustrated in Figure 69, Heck coupling of commercially available 5-iodo-2'-deoxyuridine (22) with N-allyltrifluoroacetamide provided compound 23. The5'-triphosphate group-was incorporated by treatment of compound 23 with trimethylphosphate, phosphorous oxychloride 5 (POCl 3 ), and proton sponge (1,8-bis(dimethylamino)-naphthalene) followed by tri-n butylammonium pyrophosphate, and the trifluoroacetamide group then removed with, aqueous ammonia to afford C5-modified uridine intermediate 7. {0.5051, C7-modified 7-deazaadeposine intermediate 13, the key intermediate for 7 deazaadenosine analogs, has been synthesized. As shown in Figure 70, 10 diethoxyethyleyanoacetate 24 was synthesized from bromoacetal 25 and ethyl cyanbacetate 26 following a known protocol (Davoll (1960) J. AM. CHEM. Soc. 82: 131-138). Condensation of 24 with thiourea provided pyrimidine 27, which was desulfurized with Raney nickel and then cyclized to pyrrolopyrimidine 28 with dilute aqueous HCL Treatment of 28 with POC13 afforded 4-chloro-7-deazaadenine 29. The aryl iodide group which can serve as a Sonogashira, coupling 15 partner for installation of the propargylic amine in 13 was incorporated by reacting 29 with N iodosuccinimide to generate 4-chloro-7-iodo-7-deazaadenine 30 in 13% overall yield from bromoacetal 25. Figure 71 shows glycosylation of compound 30 with protected deoxyribosyl chloride 38 (generated from deoxyribose as shown in Figure 72), followed by anmonolysis afforded 7-iodo-adenosine 39 (Gourlain et a. (2001) NUCLEiC ACIDS REs. 29: 1898-1905). Pd 20 mediated Sonogashira coupling (Seela et al. (1999) HELV. CHEM. ACTA 82:1878-1898) of 39 with N-propynyltrifluoroacetamide provides 40, which is then converted to the 5' nucleotide triphosphate and deprotected with ammonia to yield C7-modified 7-deazaadenosine intermediate 13. [05061 In order to create a library of metal-binding uridine and adenosine analogs, a 25 variety of metal-binding groups as NHS esters can be coupled to C5-modified uridine intermediate 7 and C7-modified 7-deazaadenosine intermediate 13. Exemplary metal-binding groups are shown in Figure 68 and include phosphines, thiopyridyl groups, and hemi-salen moieties. Additional deoxyadenosine derivatives, such as, for example, compounds 41 and 42 shown in Figure 73, can be prepared by coupling alkyl- and vinyl trifluoroacetamides to 8 30 brorno-deoxyadenosine (31). These intermediates then are coupled with the NHS esters shown WO 20014/016767 PCT/US2003/025984 -165 in Figure 68 to generate a variety of metal-binding. 8-functionalized deoxyadenosine triphosphates. [05071, As alternative functionalized adenine analogs that will both probe the structural requirements of DNA polymerase acceptance and provide potential metal-binding functionality, 5 six 8-modified deoxyadenosine triphosphates (Figure 74) have been synthesized. Al functional groups were installed by addition to 8-bronio-deoxyadenosine (31), which was prepared by bromination of deoxyadenosine in the presence of scandium chloride (ScCl 3 ), wlich we found to greatly iTgcrease product yield. Methyl-, (32), ethyl 2 (3), and vinyladenosine (34) were synthesized by Pd-mediated Stille coupling of the corresponding alkyl tin reagent and 31 10 (Mamos et al. (1992) TETRAHEDRON LETT. 33: 2413-2416). Methylamino- (35) (Nandanan et al. (1999) J. MED. CHEM. 42: 1625-1638), ethylamino- (36), and histaminoadenosine (37) were prepared by treatment of 23 with the corresponding amine in water or ethanol. The 5'-nucleotide triphosphates of 32-37 were synthesized as described above. Acceptance of Nucleotides by Polym erase 15 [05081 The ability of the modified nucleotide triphosphates containing metal-binding functionality shown in Figure 75 to be accepted by DNA polymerase enzymes was studied. Synthetic nucleotide triphosphates were purified by ion exchange and reverse-phase HPLC and were added to PCR reactions containing Taq DNA polymerase, three natural deoxynucleotide triphosphates, pUC19 template DNA, and two DNA primers. The primers were chosen to 20 generate PCR products ranging from 50 to 200 base pairs in length. Control PCR reactions contained the four natural deoxynucleotide triphosphates and no non-natural nucleotides. PCR reactions were analyzed by gel electrophoresis and the results indicate that functionalized uridine analogs 2, 3, 7, 13, 28, 29, and 30 were efficiently incorporated by Taq DNA polymerase over 30 PCR cycles, while uridine analogs 31 and 32 were not efficiently incorporated (see, Figure 75). 25 These results demonstrate that synthetic nucleotides containing metal-binding functionality can both be read as templates and incorporated as building blocks into non-natural nucleic acids using DNA polymerases. The 8-modified adenosine triphosphates 32 and 33 were not accepted by Taq DNA polymerase, suggesting possible rejection of modifications at CS (see, Figure 75). [05091 Functionalized nucleotides that are especially interesting yet are not compatible 30 with Taq, Pfu, or Vent thermostable DNA polymerases can be tested for their ability to participate in primer extension using other commercially available DNA polymerases including WO 2004/016767 PCT/US2003/025984 -166 the Klenow fragment of E. coli DNA polymerase I1T7 or T4 DNA polymerase, or M-MuLV reverse transriptase: Generation of Polymer Libraries [05101 Non-natural polymer libraries containing synthetic metal-binding nucleotides that 5 are compatible with DNA polymerases'have been created. Libraries of 10' 5 different r6dified nucleic acids consisting of 40 random bases flanked by two' primer binding regions and containing the imidazole-linked thymine base shown in Figure 76 have been created. These libraries were efficiently generated by three methods- stndard PCR, error-prone PCR, and primer extension using large quantities of template and stoichiometric quantities of only one 10 primer. The resulting double-stranded libraries were denatured and the desired strand isolated using the avidin-based purification system described bereinabove. Two rounds of in vitro selection oni this library for polymers that fold only in the presence of Cu 2 + have been performed using the gel elbetrophoresis selection for folded nucleic acids as described herein. [0511] Libraries of nucleic acids containing the most promising polymerase-accepted 15 metal-binding nucleotides, including 28-30 (Figure 75), can also be generated. Libraries can be generated by PCR amplification or by primer extension of a synthetic DNA template library consisting of a random region of 20 or 40 nucleotides flanked by two 15-base constant priming regions (Figure 77). The priming regions contairi restriction endonuclease cleavage sites to allow DNA sequencing of pools or individual library members. One primer contains a primary 20 amine group at its 5' terminus and will become the coding strand of the library. The other primer contains a biotinylated 5' terminus and will become the non-coding strand. The PCR reaction includes one or two non-natural metal-binding deoxyribonucleotide triphosphates, three or two natural deoxyribonucleotide triphosphates, and a DNA polymerase compatible with non natural nucleotides. Following PCR to generate the double-stranded form of the library, library 25 members then are denatured and the non-coding strands removed by washing with streptavidin linked magnetic beads to ensure that no biotinylated strands remain in the library. Libraries of up to 101 ' different members can be generated by this method, far exceeding the combined diversity of previously reported combinatorial metal-binding catalyst discovery efforts. (0512] Each library then is incubated in aqueous solution with a metal of interest from 30 the following non-limiting list of water compatible metal salts: ScCl 3 , CrC13, MnC12, FeCl 2 , FeCl3, CoC1 2 , NiCl 2 , CuC 2 , ZnC1 2 , GaC1 3 , YC1 3 , RuC 3 , RhCl 3 , Na2PdCl 4 , AgCl, CdCI 2 , InCl3, WO 2004/016767 PCT/US2003/025984 -167 SnCli, La(OTf)3, Ce(OTf) 3 , Pr(OTf) 3 , Nd(OTf) 3 , .Sm(OTf) 3 , Eu(OTf)3, Gd(OTf) 3 , Tb(OTf) 3 ,. Dy(OTf) 3 , Ho(OTf) 3 , Er(OTf) 3 , Tm(OTf)3, Yb(OTf) 3 , Lu(OTf) 3 , 1C1 3 , PtCl 2 , AuCI, HgC1 2 , HgC1,'PbC12, and BiCl 3 (Kobayashi et at (1998) L AM. CHEM. SoC. 120: 8287-8288; Fringuelli et at. (2001) EUR. J. ORo. CHEM. 2001: 439-455). The metals are chosen in part based on the 5 specific chemical reactions to be catalyzed. For example, libraries aimed at reactions such as aldol condensations or hetero Diels-Alder reactions that are known to be catalyzed by Lewis acids are incubated with ScC1 3 or with one of the lanthanide triflates (Fringuelli et at (2001) 'supra). In other cases, metals not previously known to catalyze the transformations of interest are also used to evolve polymers with unprecedented' activity. The metal-incubated library is 10 purified away from unbound metal salts using gel filtration cartridges (available'from, for example, Princeton Separations) that separate DNA oligonucleotides 25 bases ot longer from unbound smaller reaction components. ' [05131 The ability of the polymer library (or of individual library members) to bind metals of interest is verified by treating the metalated library free of unbound metals with metal 15 staining reagents, such as dithiooxamide, dimethylglyoxime, or potassium isothiocyanate (KSCN) (Francis et al. (1998) CURR. OpIN. CHEM. BIOL. 2: 422-8) or EDTA (Zaitoun et al. (1997) 5. PHYs. CHEM. B 101: 1857-1860), that become distinctly colored in the presence of different metals. The approximate level of metal binding is measured by spectrophotometric comparison with solutions of free metals of known concentration and with solutions of positive 20 control oligonucleotides containing an EDTA group (which can be introduced using a commercially available phosphoramidite from Glen Research, Sterling, Virginia, USA). Selecting Nucleic Acid Polymers [0514] Once the libraries of functionalized DNAs are synthesized and characterized, they are subjected to three types of in vitro selections for: (i) folding, (ii) target binding, or (iii) 25 catalysis. {05151 (i) Folding. Non-denaturing gel electrophoresis can be used as a simple selection, to be applied to inventive libraries of modified nucleic acids, to select for nucleic acid folding in the presence of specific metals of interest. In order to test this selection approach on molecules similar to future library members, three 60-base DNA oligonucleotides known 30 (Schultze et al. (1994) . MOL. BIoL. 235: 1532-1547) or predicted (SantaLucia (1998) PROC. NATL. ACAD. SeT. USA 95: 1460-1465) to have very different folded states were synthesized.

WO 2004/016767 PCT/US2003/025984 1 -168 Each oligonucleotide contained a core 30-base ,sequence flanked by two 15-base primer binding sequences.- The unstructured control oligonucleotide contained a poly T core and an EcoR I restriction site. The Acond core sequence contained perfect inverted repeat predicted to form a highly stable hairpin, while the third core sequence contained a poly G core known to fold in 5 solution intO an intramolecular G-quaftet (Cheng et al. (1997) GENE 197:- 253-260). Thq three DNA sequences were combined in equimolar ratios and the mixture subjected to preparatiye non-denaturing gel electrophoresis. The highCnidbility'potion of the DNA was captured and compared by inalyt c electrophoresis to authentic poly , hairpin, and poly C oligonucleotides. The results indicate that folded DNA sequences tqn be readily separated from a mixture of 10 folded and unfolded DNA'molecules by non-denaturing gel electrophoresis. This selection approach can be applied to the metal-binding polymer libraries, wherein polymers with anticipated metal binding ability will be incubated with one or more water-compatible metal sources prior to.selection. Polymers capable of folding in the presence, but not in the absence, 6f metals will serve as especially attractive starting points for the next two types of selections. 15 [05161 (ii) Target Binding. Selections for target binding can be performed by incubating the solution-phase polymer library with either immobilized target or with biotinylated target followed by streptavidin-linked beads. Non-binders are removed by washing, and polymers with desired binding properties are eluted by chemical enaturation or by adding excess authentic free ligand. In order to complete one cycle of functionalized DNA evolution, the DNA templates are 20 amplified by PCR using one primer containing the 5'-functionalized hairpin primer and a biotinylated second primeroptionally diversified by error-prone PCR (Caldwell (1992) PCR METHODS APPIc. 2: 28-33) or by nonhomologous random recombination method, and then denatured into single stranded DNA and washed with streptavidin beads to remove the non coding template strand. The resulting pool of selected single-stranded, 5'-functionalized DNA 25 completes the evolution cycle and enters subsequent rounds of DNA-templated translation, selection, diversification, and amplification. [05171 (iii) Catalysis. Selection for synthetic polymers that catalyze bond-forming or bond-cleaving reactions can also be performed. Library members that catalyze virtually any reaction that causes bond formation between two substrate molecules or that results in bond 30 breakage into two product molecules can be selected using the schemes proposed in Figures 12 and 13. As illustrated in Figure 12, in order to select for bond forning catalysts (for example, WO 2004/016767 PCT/US2003/025984 -169 hetero Diels-Alder, Heck coupling, aldolreaction, orolefin metathesis catalysts), library members are covalently linked to one substrate through their 53 amino or thiol termini. The other substrate of the reaction is synthesized as a derivative linked to biotin. When dilute' solutions of library-substrate conjugate are reacted with the substrate-biotin conjugate, those 5 library members that catalyze bond formation cause the biotin group to become covalently attached to themselves. Active bond forming catalysts can then be separated froi inactive library members by capturing the former with immobilized streptavidin and washing away inactive polymers. By way of example, the synthesis and selection of active Heck couplingg catalysts, active hetero diels-alder catalysts and active aldol addition catalysts may be performed 0 as shown in Figures 78A, 78B, and 78C, respectively. [0518] In an analogous manner, library members that catalyze bond cleavage reactions such as retro-aldol reactions, aide hydrolysis, elimination reactions, or olefin dihydroxylation followed by periodate cleavage can also be selected, as illustrated in Figure 13. In this case, . metalated library members are covalently linked to biotinylated substrates such that the bond 5 breakage reaction causes the disconnection of the biotin moiety from the library members. Upon incubation under reaction conditions, active catalysts, but not inactive library members, induce the loss of their biotin groups. Streptavidin-linked beads can then be used to capture inactive polymers, while active catalysts are able to elute from the beads. Related bond formation and bond cleavage selections have been used successfully in catalytic RNA and DNA evolution 0 (Jaschke et al (2000) CURR. OPIN. CiM. BIot. 4: 257-62). Although these selections do not explicitly select for multiple turnover catalysis, RNAs and DNAs selected in this manner have in general proven to be multiple turnover catalysts when separated from their substrate moieties (Jaschke et aL (2000) CuRR. OmN. CH-M. BIOL. 4: 257-62; Jaeger et al (1999) PROc. NATL. ACAD. SC. USA 96: 14712-7; Bartel et al (1993) SCIENCE 261: 1411-8; Sen et al. (1998) CURR. 25 OPIN. CHEM. BIOL. 2: 680-7). [05191 It is contemplated that catalysts of three important and diverse bond-forming reactions (Heck coupling, hetero Diels-Alder cycloaddition, and aldol addition) can be created using the technologies described herein. All three reactions are water compatible (Kobayashi et at (1998) J. AM. CHEM. SOC. 120: 8287-8288; Fringuelli et al (2001) EUR. J. ORG. CHEM. 2001: 30 439-455; Li et al (1997) ORGANIC REACTIONS IN AQUEOUS MEDIA) and are known to be catalyzed by metals.

WO 2004/016767 PCT/US2003/025984 -170 Evolving Functionalized DNA Polymers 105201 Following each rouid of sectionn. activlibrary members can be amplified directly by PCR with the non-naturai hucledtides and subjected to additioiiaf rounds of selection to enrich the library for desired catalysts. Libraries may be diversified by random mutagenesis 5 using errorprone PCR or by nonhomologous recombination and characterized by DNA sequencing before and after selection: Because error-prone PQR is inherently less effidient than normal PCR, error-prone PCR diversificatiohii conducted with only natural nucleotides. The mutagenized DNA templates then are translated into non-natural nucleic acid polymers as described aboye. 10 [05211 In addition to simply evolving. ctive catalysts, the in vitro selections described herein may be used to evolve catalysts withproperties difficult to achieve using current catalyst discovery approaches. For example, substrate specificity among catalysts can be evolved by selecting for active catalysts in the presence of the desired substrate and thdn selecting for inactive catalysts in the presence of one or rhdre undesired substrates. Using this'strategy, it is 15 contemplated that it will be possible to evolve libraries of catalysts with unprecedented regio and stereoselectivity. By way of example, four types of substrate specificity currently unachievable by known catalysts nor likely to be solvable by current catalyst discovery methods include: (i) Heck catalysts that operate on para- but not meta- aryl chlorides, (ii) aldol catalysts that accept ketones but not aldehydes as enolate acceptors, (iii) hetero Diels-Alder catalysts that 20 reject olefin dienophiles, and (iv) hetero Diels-Alder catalysts that accept trans-trans but reject cis-trans or terminal dienes. Metal-binding polymers containing well-ordered, three-dimensional dispositions of key steric and electronic groups may be ideally suited to solving these problems. Similarly, metal selectivity can be evolved by selecting for active catalysts in the presence of desired metals and selecting against activity in the presence of undesired metals. Catalysts with 25 broad substrate tolerance may be evolved by varying substrate structures between successive rounds of selection. Characterizing catalysts evolved by the above methods may provide new insights into developing analogous small molecule catalysts with powerful and unprecedented selectivities. [0522] In addition, the observations of sequence-specific DNA-templated synthesis in 30 DMF and CH 2 Cl 2 suggested that DNA-tetralkylammonium cation complexes may form base paired structures in organic solvents. These findings raise the possibility of evolving non-natural WO 2004/0)16767 PCT/US2003/025984 -- 171 nucleic acid catalysts in organic solvents using slightly modified versions of the selections described above. The actual bond forming and bond cleavage selection reactions iay b& conducted in organic solvents, the crude reactions then will be ethanol precipitated to, remove the tetraalkylammonium cations, and the immobilized avidjri separation of biotinylated and non 5 biotinylated library inembers in aqueous, solution will be performed. PCR amplification of selected members will then take place as. deschbed hereinabove. Successful evolution 6f reaction catalysts that function in organic solveits would expand considerably both the scope of reactions'that can be catalyzed and the utility 6f the resulting evolved non-natural polymer catalysts. 10 Example 11: In Vitro Selection for Protein Binding and Affiiy 105231 This Example demonstrates that it is possible to perform in vitro selections for nucleic acid-linked synthetic small molecules~with protein binding affinity. These selections (i) offer much greater sensitivities (1020 mol) than previously reported synthetic molecule screens for protein binding, (ii) can be rapidly iterated to achieve >I 0 6 -fold net enrichments of active 15 molecules, and'(iii) can be adapted to select for binding specificity. 10524] Because all molecules in a selection are processed simultaneously, selections offer much higher potential throughput than screens. Selections typically do not require sophisticated equipment and can be iterated to multiply the net enrichment of desired molecules. Certain properties such as binding specificity, although difficult to screen, can be readily 20 selected. Finally, the outcomes of laboratory and natural selections usually are linked to amplifiable nucleic acids, permitting the selections to offer far greater sensitivities than screens. The covalent linkage of oligonucleotides to corresponding synthetic molecules, either as a consequence of nucleic acid-templated organic synthesis or as a result of conjugating a nucleic acid to synthetic molecules, allows synthetic molecules to be selected and then identified. 25 Despite these attractions, selections for synthetic molecules have been largely unexplored. {0525] At the outset, a variety of synthetic small molecules conjugated to 36- to 42-base DNA oligonucleotides (see, Figure 79) were synthesized such that each small molecule was linked to a unique DNA sequence. The small molecules were chosen either for their known binding affinities to six proteins (see, Figure 79), or as nonbinding negative controls. Solutions 30 containing mixtures of DNA-linked protein ligands and DNA-linked negative controls were used WO 2004/016767 PCT/US2003/025984 - 172 to simulate DNA-templated synthetic small molecule libraries containing small fractions of library members with protein binding activities. 105261] Selections for protein affinity were performed by incubating mixtures of DNA linked synthetic small molecules for 1-2 hours with target ,proteins covalently conjugated to 5, beads. The non-binders were removed by washing the beads with high salt buffer. The bound molecules were then PCR amplified to amplify the DNA oligonucleotides surviving selection. Sequences encoding known protein binding ligands were distinguished from DNA encoding. non-binders by digestion with sequence-specific restriciion endonucleases, permitting their relative ratio to be quantitated by gel electrophoresis and densitometry. The efficiency of each 10 selection was assessed by the degree to which DNA- linked protein ligands were enriched relative to DNA-linked non-binders (the "enrichment factor"). [0527] Among the protein-small molecule interactions considered, the binding of glutathione amide to glutathione S-transferase (GST) is among the lowest affinity (Kd = -10 pM) and, therefore, represents a stringent test of protein binding selections for DNA-linked synthetic 15 small molecules. To measure the sensitivity and efficiency of these selections (see, Figure 80), the number of DNA-linked glutathione molecules (1) were varied from 103 to 107 molecules. A 100- to 106-fold molar excess of the negative control N-formyl-Met-Leu-Phe linked DNA (2) was combined with (1) and the resulting mixture was selected for binding to GST-linked agarose beads. The selection strongly enriched as few as 10,000 copies of the DNA-linked glutathione 20 by 100- to >10 4 -fold relative to the negative control (Figure 80). Although the concentrations of DNA-linked molecules during selections were much lower than pM, the selections were successful because GST was immobilized at an effective concentration exceeding -10 pM and, therefore, permitted a significant fraction of (1) to remain bound to GST. These results demonstrate that selections for modest protein affinities (for example, Kd = -10 pM) are possible 25 in this format. [05281 In order to evaluate the generality of this approach, analogous selections were performed for binding to streptavidin, carbonic anhydrase, papain, trypsin, and chymotrypsin in addition to GST (Figure 79). Collectively these six functionally diverse proteins bind the ligands shown in Figure 79 with predicted affinities that span more than eight orders of 30 magnitude (KEd ~4 piM to -40 fiM) (D'Silva (1990) BiOCHEM, L 271: 161-165) (Jain et al. (1994) J. MED. CHEM. 37: 2100-2105; Green (1990) METHODS ENZ. 184: 51-67; Otto et al.

WO 2004/016767 PCT/US2003/025984 -173 (1997) CHEM. REV; 97: 133-172). 'In each of these ases, selection enriched < 10-16 Ml of a known smallmolecule ligand conjugated to DNA by at least 50-fold over a nop-biiiding iegativ control (Figure 79); indicating that DNA codjugation does-not impair the ability of the ligands in Figure 79 to bind their cognate protein targets and suggesting that these'selections may be 5 applicable to a wide variety. of unrelated proteins. 10529] Firtherrpore,'selections canbe iterated to itnltiply the net enrithment of desired, ''I II molecules. To test this possibility with DNA'lirked synthetic molecules, a 1:1,000 mixture. of . DNA-linked phenyl sulf6namide (3):DNA-linked N-fopylMet-Leu-Phe (2) was subjected to a selection for binding carbonic anhydrase. The molecules surviving the first selection were eluded 10 and directly subjected to a second selection usirg fresh immobilized carbonic anhydrase. PCR amplification and restriction digestion reveal6d-that the first round of selection yielded a 1:3 ratio of (3):(2), representing a 330-fold enrichment for the I5NA-linked phenyl sulfonami4e. The second round of election further enriched 3 by more than 30-fold, such that 'the ratio of (3):(2) following two rounds of selection exceeded 10:1 (>10 4 -fold net enrichment). Similarly, three 15 rounds of iterated selection were used to enrich a 1:106 starting ratio of (3):DNA-linked biotin (4) by a factor of 5 x 10 into a solution containing, predominantly DNA-linked phenyl sulfonamide (3) (see, Figure 81). These findings demonstate that enormous net enrichments for DNA-linked synthetic molecules can be achieved.through iterated selection, and suggest that desired molecules represented as rarely as 1 part in 106 (approximately the largest number of 20 different small molecules generated in a single library to date) within DNA-templated synthetic libraries may be efficiently isolated in this manner. [0530] In addition to binding affinity, binding specificity is a broadly important property of synthetic molecules. Library screening methods for binding specificity typically require duplicating the entire screen for each target or non-target of interest. In contrast, selections for 25 specificity in principle can be performed in a single experiment by selecting for target binding as well as for the inability to bind one or more non-targets. In order to validate selections for specificity among DNA-linked synthetic small molecules, DNA-linked biotin (4), DNA-linked chymostatin (5), and DNA-linked antipain (6) were combined into a single solution in a 24:4:1 ratio, respectively. Because biotin has no significant affinity for chymotrypsin or papain, 30 chymostatin binds to both proteases, and antipain binds only to papain, (see, Figure 82) this WO 2004/016767 PCT/US20031025984 -174 mixture simulates a library containing predominantly nonbinding molecules with a minor fraction of nonspecific binders and an even smaller fraction of a target-specific binder. 105311, When this mixture was subjected to two rounds of selection for binding to papain, both 5 and 6 were enriched at the expense of 4, as expected (Figure 82). However, when the 5 above mixture was washed with chynotrypsin-linked beads and selected for binding to papain in the presence of excess free chymotrypsin, only the papain-specific ligand (6) was enriched (Figure 82). The ability of the selections described Above to separate target-spetific and nox specific DNA-linked synthetic molecules from a single solution suggests their use to discover synthetic molecules that exclusively bind a single member of a large family of related proteins 10 (e.g., kinases, proteases, or glycotransferases), and that do not bind proteins that commonly reduce the biological efficacy of small molecules (e.g. by sequestering, exporting, or metabolizing them). 105321 In summary, this Example demonstrates the feasibility of performing in vitro selections for DNA-linked synthetic small molecules with protein binding activities. 'The 15 application of methods developed here to nucleic acid-templated (or nucleic acid-conjugated) libraries may play an important role in the discovery of synthetic molecules with desired properties using powerful selection and amplification strategies previously available only to biological molecules. Materials and Methods 20 DNA Synthesis [0533] DNA oligonucleotides were synthesized on a PerSeptive Biosystems Expedite 8090 DNA synthesizer using standard phosphoramidite protocols. All reagents were purchased from Glen Research, Sterling, Virginia, USA. The templates for the glutathione S-transferase (GST) selection were synthesized using a 5'-amino-modifier C12 and all other templates were 25 synthesized using 5'-amino-modifier C5. Preparation of Compound (1) 105341 Glutathione was synthesized on the solid phase using standard Boc chemistry at room temperature. 200 mg PAM Resin (Advanced ChemTech) was swelled in 2 mL DMF for 20 minutes. N-Boc-glycine (Sigma, 640 pmol, 112 mg), diisopropylcarbodiimfide (570 pmol, 89 30 pL), and 4-dimethylaminopyridine (DMAP, 57 pmol, 7 mg) were added to the resin and stirred for 4 hours. The resin was washed with DMF and then with DMF/CH 2 C1 2 (1:1). The N-Boc WO 2004/016767 PCT/US2003/025984 175., protecting group was removed using two 3 minute ashes of trifluoroacetic acid (TFA):m-cresol (95:5). The 'resin then was washed ivitb'DiMF:ClH 2 C' (1:1) and DMF;pyridine (I). ,A'solutioni of N-Boc-Cys(PP YH(CeMImpex, 800 pinol, 320 ing), O-(7-Azabenzot'iazol-1-yl) N,N,NU'-tetramethyluIonium hexafluorophosphate (Aldrich, 720 rnol, 274 mg), 2,6-lutidine 5 (1.2 mmol, 131 pl) and YN-diisopropylethylamflin6 (DIPEA, 750 pmol, 131 pl) in 800 pL of 1 methyl-2pyrroli'dinone was stirred for 15. minutes and' then 'added to the resin; stirring for 30 minutes. Thp resin then was washed with D1F/CH 1 :1). To remove the N-Boc protecting, group on cysteine, a solution of trimethylsilyl triflate (TMS-Otf) (2.8 m1, 0.5 ml) and 2,6, lutidine (4.58 mmol, 0.5 mL) in 1.75 mL CH 2 C1 2 'was added to the resin and stirred for I hoar. 10 The resin then was washed with methanol and then with D MF:CH 2

C

2 (1:1). Fmoc-Glu-OFm (Chemlmpex, 800 pmol, 438 mg) was coupled as described above. The fully protected glutathione was cleaved from the resin with a solution'of trifluoromethanesulfonic acid:m cresol:thioanis6le:TFA (2:1:1:8), stirring for 1 hours. The mixture was filtered and the filtrate was extracted into hexane. The crude extract Was purified-using preparative thin layer 15 chromatography in hexane. The silica containing the crude product (Rr = 0.35) was washed extensively with hexane:ethyl acetate (4:1). The filtrate was isolated under vacuum to afford a yellowish solid. Yields for this synthesis were not optimized. [0535] A solution of protected glutathione (1.1 minol, 4mg) in 90 1d DMF with N hydroxysuccinimide (NHS, 11 pmol, 1.3 mg), dicyclohexylcarbodiimide (DCC, 11 pamol, 2.3 20 mg), and DMA-P (5.7 ptmol, 0.7 mug) was agitated for 1 hour. The mixture was spun down and the supernatant was added to 5'-amino-terminated protected DNA on CPG beads. This mixture was agitated for 2 hours and then the beads were washed with DMF, with CH 3 CN, and dried with nitrogen. Preparation of Compound (2a) 25 [05361 N-formyl-Met-Leu-Phe (MLF) was purchased from Sigma and coupled to 5' amino-terminated protected DNA on CPG beads using the conditions described for compound (1). Preparation of Compound (2b) [05371 MLF (10-100 pamol, 0.17 M) was dissolved in dry DMF with I equiv. 1 30 hydroxybenzotriazole (Novabiochem), 0,9 equiv. O-Benzotriazol-1-yl-NNN',N' tetramethyluronium hexafluorophosphate (Aldrich), and 2.3 equivalents of DIPEA. The solution WO 2004/016767 PCTIUS2003/025984 -176 wasagitated at room temperature for 1 hour and then added to a unique sequence of 5'-amino tenminated protected DNA on CPG beads'. The mixture was agitated for 1 hour at room temperature. The beads then were washed with DMF, thek with CH 3 CN, and dried undet nitrogen. 5' Preparation of Compound (3) [0538) .Fmoc-Lys(Mmt)-OH (Novabiochem) was attached to amino-terminated protected DNA on CPG beads using the method described for compound (2b). The Fmoc group was i, enoved yith three 2 minute washes with 20% piperidine in DMF. The mixture then was washed with DMF and then with CH 3 CN. The a-amine then was capped with a solution of 5% 10 1 -methylimidazole in acetic anhydride/pyridine/tetrahydrofuran (1: 1.1:18) for 10 minutes at room temperature. The beads then were washed with DMF and CH 3 CN, and then treated with 3% trichloroacetic acid, 1% thioanisole in CH 2 Cl 2 for 5 minutes at room temperature to remove the Mmt protecting group. The mixture was washed 'vith CH 3 CN and dried with nitrogen. Fmoc-Phg-OH (Novabiochem) was attached to the s-amine of the Lys-linked DNA using the 15 method described for compound (2b). After removal of the Fmoc protecting group, 4 carboxybenzenesulfonamide (Aldrich) was attached to the beads using the method described for compound (2b). The beads were washed with DMF, then with CH3CN, and dried with nitrogen. Preparation of Compounds (4a, 4b) [05391 A 5'-biotin modified phosphoramidite (Glen Research, Sterling, Virginia, USA) 20 was used as the final monomer in the DNA synthesis. Preparation of Compound (5) [0540] Chymostatin (Sigma) was attached to amino-terminated protected DNA on CPG beads using the conditions described for compound (2b). Preparation of Compound (6) 25 10541) Antipain (Sigma, 1.5 jimol, 0.9 mg) was added to a 30 pLL solution of 300 mM DCC and 300 mM NHS in DMF. After agitating for 1 hour at room temperature, this solution was added to 45 ptL of 5'-amino terminated DNA (-200-300 pM) in 0.1 M MES buffer pH 6.0. This DNA had previously been cleaved from the CPG beads and purified by HPLC as described in the next section. After 2 hours, this solution was purified by gel filtration using Sephadex G 30 25 followed by reverse-phase HPLC.

WO 2004/016767 PCT/US2003/025984 -177 [0542] The complete structures of synthetic groups 1 r6 linked to DNA are shown in Figure 83. Characterization of DNA-linked Synthetic Molecules 105431 Small molecule-DNA conjugates were cleaved ftom the CPG beads with a 5 solution of methylamine:ammonium hydroxide (1:1) at 55 "C for 1 hour. The solution was dried under vacuum and then purified by reverse phase HPLC using TEAA/CH 3 CN gradient and analyzed by MALDI-TOF mass spectrometry. Stock solution concentrations were determined using UV,-Vis spectroscopy and serial dilutions wer' prepared for the selection experiments. Samples were stored in water at -20 "C. 10 Preparation of Immobilized Target Proteins [0544] NIS-activated Sepharose 4 Fast Flow (Amersham Pharmacia) was prepared in accordance with the manufacturer's instructions. Equine (GST, bovine carbonic anhydrase (CA), papain, Na-p-tosyl-L-lysine chloromethyl ketone (TLCK)-treated bovine chymotrypsin, and N p-tosyl-L-phenylalanine chloromethyl ketone (TPCK)-treated bovine trypsin were purchased 15 from Sigma. Typically, proteins were dissolved in phosphate buffered saline (PBS) buffer pH 7.4-7.6 at concentrations of 20-100 pM. Protein concentrations were determined using UV-Vis spectrometry. Proteins were incubated with beads for 16 hours at 4 *C. The beads were capped for two hours with Tris buffer, then washed extensively with the appropriate selection buffer containing 1 M NaCl and then exchanged into the appropriate selection buffer (see, Table 14). 20 Beads were stored for up to 1 month at 4 "C in a volume of selection buffer equal to the initial volume of beads used. Before use, papain beads were activated using a solution of 5.5 mM cysteine HCi, 1.1 mM EDTA, and 0.067 mM p-mercaptoethanol for 30 minutes at 4 "C. Streptavidin magnetic particles (Roche) were washed 3x with selection buffer before use. TABLE 14: Selection and Wash Buffers Protein Composition of Seletion Buffer Composition of Wash Buffers GST PBS pH 7.4 Carbonic 10mM Tris pH 7.4, 0.1 M NaCI 10 mM Tris pH 7A, 0.25-0.5 M NaCI Anhydrase Papain 50 mM Tris pH 7.4, 0.1 M NaC, 1 50 mM Tris pH 7.4, 0.5 M NaCl, 1 mM mM EDTA EDTA WO 2004/016767 PCTUS2003/025984 '- 178 Trypsin 50 mM Tris pH 8:0, 0.1 M NaCl" 10' 50-mM Tris pH 8.0, 0.5 M NaC1, 10 mM , mM CaC12 CaC12 Chymotrypsin 50 MM Tris pH 8.0, 0.1 M:NaC, 10 150 mM Tris pH 8.0, 0. 5 M NaCl; 10 mM . mM CaCl 2 Ca0 2 Streptavidin IO mMTris p1 7.4, 0.1 M NaCli 1 10mM Tris pH 7.4, 1.O M NaCl, 1 mM mM EDTA EDTA GST Selection [0545] The amount of compound (1), the binding ligand, was varied between 103 and 107 molecules and comp6uhd (2a), the non-binding ligand, was used in l'02106 molar excess. (1) 5 and (2a) were added to 40 ptL of GST beads and agitated at 4 "C for 1 hour. The mixture was transferred to a 5.0 [mi low-binding Durapore'mnmtbrane spin filter (Millipore), washed with 2x 150 pL PBS pH 7.4, lx 100 pL 0.1 M Tris pH 8.0, 0.5 MNaCl, and lx150 LPBS Thebound ligands were eluted by agitating the beads with 100 iL 0.1 M glutathione (Sigma) at rodm temperature. The eluant was ethanol precipitated with 3 M sodium acetate and 1 iL glycogen. 10 The precipitate was used directly for PCR. Carbonic Anhydrase Selection [05461 Compound (2b), the non-binding ligand, and compound (3), the binding ligand, were added to 40 ptL of resuspended beads and were diluted to 400 pL with selection buffer. Ratios were similar to those for the GST selection. The mixture was agitated at 4 "C for 1-2 15 hours. Selections then were carried out at room temperature. Each mixture was transferred to a spin filter and washed 3x with 400 pL of wash buffer and Ix 400 pL with selection buffer. The resin was removed from the spin filter with 60 pL of selection buffer and the resulting beads were subjected to PCR. Papain Selection 20 10547] Compound (4a), the non-binding ligand, and compounds (5) or (6), the binding ligands, were incubated with papain beads and selected as described for the carbonic anhydrase selection.

WO 2004/016767 PCT/U S2003/025984 -179 Chymotrypsin Selection. 105481 Compound (4a), the non-binding ligand, and compound (5), the binding ligand, were incubated with chymotrypsin beads'and selected as aesgribed for the carbonic anthydrase selection. 5 Trypsin Selection [0549] Compound (4a), the non-binding ligand, and compound (6), the binding ligand, were incubated with trypsin beads and selected as described for carbonic anhydrase. Streptavidin Selection [0550] Compound (3), the non-binding ligand, and compound (4b), the binding ligand, 10 were incubated with 15 pL streptavidin magnetic particles and agitated at room temperature for 20 minutes. Using a MPC-S magnet (Dynal), the beads were washed 2x with 0.1 M NaOH, 1 mM EDTA (100-200 p.L), 4x with wash buffer (100-200 pL), and 1x with selection buffer. The beads then were resuspended in 15 ptL double distilled H 2 0. Iterated Carbonic A nhydrase Selection 15 [0551] 108 molecules of compound (3) and 10" molecules of compound (2b) were incubated with 40 tiL carbonic anhydrase beads for 1 hours and then selected as described. After the first round of selection, 5 jiL of resuspended agarose beads were removed for PCR. 6 M guanidinium lC1, 10 mM EDTA (40 pL) was added to the beads and the mixture was heated to 90 "C for 15 minutes. The beads were filtered away using a Wizard Minicolumn (Promega). 20 The filtrate was buffer exchanged into selection buffer using a Centrisep Spin Column (Princeton Separations). A new aliquot of carbonic anhydrase beads was added to the eluted templates. After a second round of selection, the agarose beads were suspended in 30 kL of H 2 0 and 15 piL were used for PCR, The PCR products were digested with Hind Il, generating the results in Figure 84. 25 [05521 The triple iteration selection was carried out essentially as described above with a few minor changes. The prepared carbonic anhydrase beads were incubated with ZnSO4 (1 mM) for 1 hour and then washed extensively with selection buffer containing 2 M NaCL The beads were exchanged back into selection buffer and used directly for the iterated selection. I0 molecules of compound (3) and 1015 molecules of compound (4b) were added to the beads and 30 selected as described above. After the first round of selection, 3 pL aliquot was removed for WO 2004/016767 PCT/US2003/025984 PCR. A second round of selection was carried dut s described above and 8 p.L aliquot of beads was removed for PCR. After a third round 6f sele&tion, the resulting beads were removed from the spin filter using 0 L of double distilled H20 and 15 pLL of resuspended beads were used for PCR. 5 Papain Affinity And Papain Specificity Selections [05531 Affinity selection: 6x10 9 molecuves of compound (6), 2.3x1 0 molecules compoundJ (5), and .1.4x10i molecules of compound (4a) were added to 40y.L papain beads for' 1 hour. The beds were washed with papairi wivash biffy (3 x 100 gL) and once with 100 iL papain selection buffer. The beads were removed from the spin filter with 30 pL of double 10 distilled H20. A 3 pL aliquot of resuspended beads were removed for PCR, The DNA conjugates were eluted from the beads by adding 70 ptL 6 M guanidinium HCl and heating the mixture to 90*C for 15 minutes. The eluted material was buffer exchanged as described in the iterated carborfi6 anhydrase selection. After a second round of selection, the agarose beads were removed from the spin filter using 30 ptL H20 and 15 pL of resuspended beads were used for 15 PCR. [05541 Specificity selection: The same amounts of antipain, chymostatin, and biotin were added to 40 ptL chymotrypsin agarose beads in chymotrypsin selection buffer and incubated for 1 hour. The beads were spun down and the flow through was added, to 40 L fresh chymotrypsin beads and incubated for 1 hour. The beads were spun down and 15 pL of 100 pM 20 chymotrypsin in papain selection buffer was added to the flow through and then incubated for 1 hour. This solution was added to 40 iL of papain beads and selected as described above. The small molecule-DNA conjugates were eluted and buffer exchanged as described, incubated with 15 ±L 100 pM chymotrypsin for 1 hour and then subjected to a second round of selection. The beads were removed from the spin filter with 30 pL of H20 and 15 piL were used for PCR. 25 Contamination Controls {05551 Due to the high sensitivity of these experiments, two important contamination controls were used throughout these studies. First, each selection was carried out as described above except no ligand-DNA conjugates were added to the protein-linked beads, which permitting testing for buffer contamination and any cross-contamination among samples.

WO 2004/016767 PCT/US2003/025984 - 181 Secondly, a PCR reaction in which no material from the selection was added was used to test for. contamination in primers, dNTPs, and PCR buffers. PCR Conditions and Gel Electrophoresis Analysis (0556] Templates surviving the selection were amplified using PCR, All reactions 5 contained I pM of each primer and 250 p.M of each dNTP (Promega). For the GST selection, the precipitated DNA was ised in the PCR reaction hnd amplified with Platinum Taq (Invitrogen). PCR conditions were step 1: 94"C, 2%; step 2: 94"C, 30 s; step 3: 55"C, 1'; step 4:72 0 C, 3.0 s; step 5: go to step 2, x29; step 6: 72"C,:5'; step 7: hold at 4"C. For all other selections, the agarose beads (3-15 pL) were used directly in the PCR reaction with Taq 10 polymerase (Promega). PCR conditions were step 1:.94"C, 2' step 2: 94"C, 30 s; step 3: 55C, l'; step 4: 72"C, 30 s; step 5: go to step 2, x24; step 6: 4"C. [0557] The PCR products then were digested for 1-2 hours with the restriction enzymes (New England Biolabs, 5-10 units) that digest the ligand-encoding DNA. Digestion products were analyzed by electrophoresis on 3% agarose gels and quantitiated by ethidium bromide 15 staining and densitometry on a Strategene Eagle Eye II system. Enrichment Calculations [0558] Enrichment ratios are calculated as the ratio of the fraction of binding ligand surviving the selection as determined by restriction digestion to the fraction of binding ligAnd entering the selection as determined by the known concentrations of the stock solutions. 20 DNA Sequences of Templates and Primers 10559] Restriction endonuclease cleavage sites are underlined. DNA Sequences for Glutathione S Transferase Selections: 105601 GSH-template (1): 5'-GCC TCT GCG ACC GTT CGG AAG CTT CGC GAG 25 TTG CCC AGC GCG (Hind III) [SEQ ID NO: 112] [0561] MLF-template (2a): 5'-GCC TCT GCG ACC GTT CGG GAA TTC CGC GAG TTG CCC AGC GCG (Eeco RI) [SEQ ID NO: 113] [05621 Primer 1: 5'-GCC TCT GCG ACC GTT CGG [SEQ ID NO: 114] [0563] Primer 2: 5'-CGC GCT GGG CAA CTC GCG [SEQ ID NO: 115) WO 2004/016767 PCT/US2003/025984 -182 DNA Sequences for Carbonic Aniydrase Selections: {0564] Phenyl sulfonamide-ten'plate (3): 5'-CGA TGC TAG CGA, ACG AAG CTT CCA CTG CAC GTC.TGC (Hind III) [SEQ ID No: 11,6 5 (05651 MLF-templdte (2b): 5'-CA TGC tAG. GA AGG GAA TTC CCA CTG CAC GTC TGC (EcoRI) [SQ ID NO: 1173 [05661 , Biotin-template (4b):. 5'-4tGA T TCAG CGA AGG GAA TTC CCA CTGCAC GTC TGC (Eco.RI) [SEQ ID NO: 118] [0567] 'Primer 1: 5'CGA TGC TAG CGA AGG [SEQ ID NO: 119] 10 {05681 Primer 2: 5'-GCA GAC GTG CAG'TGG -[SEQ ID NO: 120] DNA Sequences for Protease Selections: [0569] Chymostatin-template (5): 5'-GCA GTC GAC TCG ACC GGA TCC GGC TAC GAC GTG CAC (BaM HI) [SEQ ID NO: 121] 15 (05701 Antipain-template (6): 5'-GCA GTC GAC TCG ACC CAG CTG GGC TAC GAC GTG CAC (Pvu II) [SEQ ID NO: 122) [05711 Biotin-template (4a): 5-GCA GTC GAC TCG ACC AAG CTT GGC TAC GAC GTG CAC (Hind III) [SEQ ID NO: 123] [05721 Primer 1: 5'-GCA GTC GAC TCG ACC [SEQ lID NO: 124] 20 105731 Primer 2: 5'-GTG CAC GTC GTA GCC. [SEQ ID NO: 125] Example 12: Identification of New Chemical Reactions [0574] This Example demonstrates that it is possible to identify the existence of new chemical reactions via nucleic acid-templated synthesis. New chemical reactions have been 25 identified as a result of experiments to select for, and characterize, bond forming reactions. [05751 A one-pot selection scheme to identify new bond forming reactions is summarized in Figure 85. Briefly, when n pool A reactants and combined with m pool B biotinylated reactants, n x m possible reaction combinations are available. When the templated reaction is performed under a particular set of reaction conditions certain combinations of the WO 2004/016767 PCT/IUS2003/025984 -183 template (e.g., reactant A27) reacts with certain torhbinations of the transfer unit (e.g., the reactant biotinylated B11). The readtio6prodiets are captured by avidin linked beads. Unreacted templates"are not captured by the ;avidin and can be removed by 'washing. The avidin captured reaction product can then be amplified,' for exarnple, by PCR, And the template 5 sequenced to determine-its.codon sequence. As shown, the amplified template included a sequence tag (coding region) for reactint A27' and.a codon sequence (anneal'ng region)for reactant B11. [0576] Figure 86 provides a schematic overvi yw of a scheme for producing a library of compounds, members of which were created by new identified chemical reactions. In order to 10 select for bond-fonnig reactions, four pool A reactants presenting either a phenyl group (AIB1 and A12) or a primary amine (A2B1 and A2B2) and two biotinylated pool B reactants presenting either a carboxylic acid (1) ora methyl ester (B2) were prepared. The two coding and two annealingregions contained different restriction digestion sites to permit the relative quantitiation-of each of the four pool A members from within a mixture. All six reactants.(250 15 mol of each pool A reactant and 500 fnol of each of BI and B2) were combined in a single pot either in the presence or absence of DMT-MM, which is known to mediate amide formation between amines and carboxylic acids (Gartner et al. (2002) AGNEW, CHEM. INT. ED. 41: 1796 1800; Kunishima et aL (2002) TETRAHEDRON 57: 1551-1558). The crude reactions were passed over streptavidin-linked magnetic beads to select for templates encoding bond-forning reactions 20 and washed with denaturant to remove pool A members that did not undergo bond formation with a pool B member. The selected molecules were eluted with free biotin and formamide. A fraction of the eluant corresponding to 5 fmol of initial-total reactants was amplified by PCR and subjected to DNA sequencing and restriction digestion to determine the ratio of the four possible reaction-encoding sequences (Le., reaction of the phenyl group with the carboxylic acid, reaction 25 of the phenyl group with the ester, reaction of the amine group with the carboxylic acid, and reaction of the amine group with the ester) (Figure 86). [0577] Combining the reactants in the absence of DMT-MM resulted in very little PCR product formation following selection. In contrast, strong PCR product was observed when the reactants were combined in the presence of DMT-MM (Figure 86), consistent with the 30 effectiveness of capturing reacted pool A members and the thoroughness of the washing steps. This result suggests that the yield of PCR product following selection for bond-forming reactions WU 2004/016767 PCT/IUS2003/025984 -184 can serve as a simple screen for the presence of bond formation within a pool of reactants. To determine the identity of the bond-forming reactants, the PCR. products were digested with Mse I, which cleaves the coding region for A2 but not Al, and Tsp45 1, which cleaves the annealing region for B2 but not B 1. An analysis of the digestion fragments revealed that reaction in the 5 absence of OMT-MM followed by selection resulted in a mixture of all four possible reaction encoding pool A members (Figure 86). In contrast, reaction in the presence of DMT-MM followed by selection generated the A2B1 sequence and no significant amount of the other three sequences (Figure 86), indicating strong enrichment for the DNA encoding bond formation between the amine and the carboxylic acid. DNA sequencing of the selected PCR products was 10 consistent with the restriction digestion analysis. These results validate the basic principle of the proposed method and system for discovering new reactions. [0578J In order to test the ability of the proposed reaction discovery system to select a single reactive combination out of an eve n larger excess of unreactive combinations, the system was programmed with three reaction possibilities (amine + carboxylic acid, aide + ester, and 15 amine + ester) and combined the corresponding DNA-linked reactants in proportions that favor the unreactive combinations (amide + ester and amine + ester) by 100-fold. In the presence of amide coupling reagent DMT-MM, in vitro selection of the resulting mixture for bond-forming reactions resulted in a >1,000-fold enrichment of the template encoding bond formation between the amine and carboxylic acid. No enrichment was observed when DMT-MM was omitted. 20 This result further supports the possibility of selecting and decoding a single reactive bond forming combination from the planned 30 by 30 matrix of 900 reaction possibilities. Validation of New Reaction Discovery (Example A) [0579] This Example shows that it is indeed possible to discover new chemical reactions using DNA-templated synthesis. A 25-reaction matrix containing the DNA-linked functional 25 groups shown in Figure 87 was generated essentially as described in Figure 9 using the omega architecture, the one-pot assembly method for pool A reactants, and an optimized codon set. Among the 25 possible reactions in this set is the Huisgen 1,3-dipolar cycloaddition (Huisgen et al. (1989) PuRE APPL. CHEM. 61: 613) between an azide and an alkyne. Sharpless and co workers recently reported (Rostoutseu et al. (2002) ANGEW CHEM. INT. ED. ENGL. 41: 2596) that 30 catalytic CuSO 4 and sodium ascorbate dramatically improve the regioselectivity and efficiency of this process, permitting a robust reaction at room temperature. A reaction discovery selection WO 2004/016767 PCT/US2003/025984 -185 was performed on-a I pmol scale using this 25-readtion matrix either in the presence or the absence of CuSO4 and sodium ascorbatb. 10580] In the presence of copper and ascorbate, selection for bond-forming reactions followed by PCR amplification and sequence analysis by restriction digestion highly enriched 5 the pool A template encoding the alkyne- and azide-encbding reactants (see, Lane 2 in Figure 87B). In contest, omitting copper and ascorb ate resulted in no enrichment for the adkyne- and, azide-encoding template (see, Lane 3 in Fignre 87B). The reaction discovery selection system therefore successfully "rediscovered" the Cu(I)-nediad coupling of an alkyne and azi&. Validation of New Reaction Discovery (Example B) 10 105811 This Example shows that the reaction identified in Example A can also be identified in a 96-reaction matrix. Briefly, a 96-reaction matrix containing the DNA-linked functional groups shown in Figure 88 was generated. Pool A contained 12 reactants (Al-Al 2) and pool B contained 8 biotinylated reactants (B1-B8). When combined, 96 different reactions were possible. 15 {05821 The reactants (10 fiol each) were combined in the presence of 500 pM Cu (I) at pH 6.0. Following reaction selection and amplification, one oligonucleotide sequence was enriched. In particular, there was a 27-fold enrichment for the template encoding the reaction between reactant A2 and reactant B5. The reaction product, like Example A, appears to have resulted from a Huisgen cycloaddition reaction. In contrast, when no Cu (I) was present, there 20 was very little PCR product with no enrichment for any combination of the reactants. Validation of New Reaction Discovery (Example C) (05831 This Example shows another example that it is possible to discover new chemical reactions using nucleic acid-templated synthesis. In particular, this Example demonstrates the discovery of a novel Pd-mediated coupling reaction. 25 105841 A library of reactants were created and combined to test for the ability of nucleic acid-templated Pd-mediated coupling reactions. Two pools of reactants (see, Figure 89) were synthesized to give 12 pool A reactants (Al-A12) and 8 biotinylated pool B reactants (B1-B8). When combined, 96 different reactions were possible. The reactants (10 fmol each) were combined in the presence of 1 mM Pd(lI) at pH 7.0. Following reaction selection and 30 amplification, five oligonucleotide sequences were enriched between 10-fold and 22-fold. Analysis of the five oligonucleotide sequences revealed that reactions occurred between (i) WO 2004/016767 PCI/LS2003/025984 -186 reactant A2 and reactant B1 (ii) reactant A2 and reactant B4, .(iii) reactant A2 and reactant B8 (iv) reactant A9 and reactant B1, and (v) reactant AID and reactant B4. [05851 As an alternative to sequencing the enriched oligonucleotides, the identity of the oligonucleotide sequences attached to the reaction products were determined by microarray 5 analysis (see, Figure 90). A library of anti-sense oligonucleotides complementary to each of the templates to be included in the reaction matrix are synthesized. Then, individual antisense oligonucleotides (l' - 9' in Figure 90), complenentay to each template are immobilized at separate addressable locations of a microarray. The sequence of each anti-sense 'oligonucleotide immobilized in the microarray is known. After nucleic acid-templated synthesis, the 10 oligonucleotides attached to the resulting reaction products (for example, P1 attached to template I and product PS attached to template 8 in Figure 90) are amplified under conditions to permit incorporation of a detectable moiety, for example, a fluorphore, into the amplified template. The amplified oligonucleotides then are denatured and combined with the microarray under conditions to permit the template oligonucleotide (for example, oligonucleotide 1 and 15 oligonucleotide 8 in Figure 90) to hybridize to its immobilized, complementary oligonucleotide. After washing to remove unbound material, the microarray may then be scanned to detect a specific binding event via detection of the detectable moiety at a particular location. Based on the location of the detectable moiety and the known sequence of the complementary oligonucleotide immobilized at that location, it is possible to determine the sequence of the 20 bound template and thus the reactants that produced the reaction product. [05861 This type of microarray analysis approach was used following reactions similar to those described in Example B (96-reaction matrix with Cu (I)) and in Example C hereinabove (96-reaction matrix with Pd (11)). The microarray analysis was found to agree with the DNA sequencing results. Furthermore, the microarray analysis was found to be more direct, more 25 sensitive, and significantly faster (at least 5-fold faster) than standard sequencing methodologies. 10587] By way of example, various products of the Pd (II) mediated reactions were detected via the microarray system, the results of which are summarized in Figure 91. Figure 91 summarizes which reactants in pool A reacted with which biotinylated reactants in pool B to create a product. Figure 91 also summarizes the level of signal over background and DNA 30 templated reaction yield for each product. Of particular interest is the discovery using both sequence analysis approaches of a bond-forming reaction between DNA-linked terminal alkyne WO 2004/016767 PCT/US2003/025984 -187 A2 and DNA-linked acrylamide B8 in the presence of 1 mM Pd(II) at pH 7 (see, Figures 89 and 91). This- reaction is comparable in effitiencya DNA-templated Heck coupling reactions of aryl iodides and olefins ahd- does not proc'ed in the absence of a Pd source. Although Pdrmediated couplings between terminal alkynes and aryl iodides are known (Amatore et aL (1995) J. ORG. 5 CHEM. 60:'6829), the Pd-mediated coupling of terminal alkynes with simple dr electro deficient olefins appearsto be a new type of reaction scheme. 'This newly discovered reaction scheme may now be characterized in greater detail usitg more conventional larger scale reactions. INCORPORATION BY REFERENCE [0588] The entie contents of each of the publications, patents and patent applications' 10 cited herein are incorporated by reference into this application for all purposes. EQUIVALENTS [0589] The invention may be embodied in other specific forms without departing form the spirit or essential characteristics thereof The foregoing embodiments are therefore to-be considered in all respects illustrative rather than limiting on the invention described herei7. 15 Scope of the invention is thus indicated by the appended claims rather than by the foregoing description, and all changes that come within the meaning and range of equivalency of the, claims are intended to be embraced therein.

Claims (1)

1. 2. is captured by avidin or streptavidin coupIed to a solid support. 1 102. The method of claim 96, wherein said first reactive unit is covalently attached to 2 said first oligonucleotide, 1 103. The method of claim 96 or 102, wherein said second reactive unit is covalently 2 attached to said second oligonucleotide.